chore!: 🧹 bump version

closes #230

.github/workflows/publish_action.yml

@@ -5,14 +5,20 @@ on:
     tags:
       - '*'
 
+permissions:
+  issues: write
+
 jobs:
   publish-node:
     name: Publish Custom Node to registry
     runs-on: ubuntu-latest
+    if: ${{ github.repository_owner == 'melMass' }}
     steps:
       - name: ♻️ Check out code
         uses: actions/checkout@v4
+        with:
+          submodules: true
       - name: 📦 Publish Custom Node
-        uses: Comfy-Org/publish-node-action@main
+        uses: Comfy-Org/publish-node-action@v1
         with:
           personal_access_token: ${{ secrets.COMFY_REGISTRY_TOKEN }}

CHANGELOG.md

@@ -3,10 +3,193 @@
 This is an automated changelog based on the commits in this repository.
 
 Check the notes in the [releases](https://github.com/melMass/comfy_mtb/releases) for more information.
-## [main] - 2024-03-07
+## [main] - 2025-04-16
 
 ### Bug Fixes
 
+- 🐛 note+ breaking wfs ([af42134](https://github.com/melMass/comfy_mtb/commit/af421340286b234e4c0cfcd4143a9d8726ebf3d1))
+- 🐛 ColorCorrect clamp issue ([a85e57b](https://github.com/melMass/comfy_mtb/commit/a85e57b18c7d3c765131873ffff523244ca9be73))
+- 🐛 Whisper chunks processing ([8bf3545](https://github.com/melMass/comfy_mtb/commit/8bf3545fec5b2a180607d40394b025a1e09c14b6))
+- 🐛 stackImages move to device ([d87e52e](https://github.com/melMass/comfy_mtb/commit/d87e52ea2c112fd95f257dcd6a54a5db77a34fc3))
+- 🐛 bbox upscale from center ([55261bd](https://github.com/melMass/comfy_mtb/commit/55261bda7c33d088b62c5483e4483201e5a9ce77))
+- 🐛 add MASK support for PickFromBatch ([0d264b9](https://github.com/melMass/comfy_mtb/commit/0d264b90a78d5a6719fb3ce71f4e9a642db4c950))
+- 🐛 use addDOMWidget for Debug node ([46af602](https://github.com/melMass/comfy_mtb/commit/46af6027d6c87d0c29b8bb0fd1cc1dbdae993629))
+- 🐛 use "modern" notation in toDevice ([b7ca8ed](https://github.com/melMass/comfy_mtb/commit/b7ca8ed1c6e117b71afd7696f55dcc3dbd5bad08))
+- 🐛 handle missing submodules ([d61da30](https://github.com/melMass/comfy_mtb/commit/d61da304099ff5e4528e4beb1ecc2eb83cabaaa1))
+- 🐛 add warnings about what each IO mode can do ([6608c0b](https://github.com/melMass/comfy_mtb/commit/6608c0b6d1cf8f7a9901214096f8c78bfe17056f))
+- 🐛 soft deprecate compression h264 ([a757e1c](https://github.com/melMass/comfy_mtb/commit/a757e1c98b2abbd2221a15b77e89d772e02d1d82))
+- 🐛 limit packages allowed to be installed from API ([d6e004c](https://github.com/melMass/comfy_mtb/commit/d6e004cce2c32f8e48b868e66b89f82da4887dc3))
+- 🐛 ensure default settings (io sidebar) ([ed17fa2](https://github.com/melMass/comfy_mtb/commit/ed17fa2ef4688aadf305a6d51b32c13a0efd22d6))
+- 🐛 spawn colour picker at pointer location ([e5482ae](https://github.com/melMass/comfy_mtb/commit/e5482aee5e3de07e8f055b3edc0fccc0e0f75c14)) by [@webfiltered](https://github.com/webfiltered) in [#223](https://github.com/melMass/comfy_mtb/pull/223)
+- 🐛 i/o sidebar for custom paths ([62469a4](https://github.com/melMass/comfy_mtb/commit/62469a4dd96e32509171aad74fcae8d2bb0ec593))
+
+### Features
+
+- ⚡ add BatchFromFolder ([9618513](https://github.com/melMass/comfy_mtb/commit/96185132b83c182032e9f6e822561eb5699af517))
+- ⚡ add use_normalized to TransformBatch2D ([d4a31bf](https://github.com/melMass/comfy_mtb/commit/d4a31bf19c2863df8dfc4cb9a3cd6683304949e4))
+- [**breaking**] ⚡ add support for masks in BatchFLoatMath ([fc7ba08](https://github.com/melMass/comfy_mtb/commit/fc7ba084f6ed7880e88e28eb448ab0bd7d796824))
+- ✨ add use_normalized to TransformImage ([4516aa9](https://github.com/melMass/comfy_mtb/commit/4516aa9cb4fcb12c946999d6dcc1501cc09011a3))
+- ✨ add regex support for String Replace ([78946b0](https://github.com/melMass/comfy_mtb/commit/78946b0fa3c3cf5dfcee8c7c4c0921b722d09d1e)) by [@poetryiii](https://github.com/poetryiii) in [#233](https://github.com/melMass/comfy_mtb/pull/233)
+- ✨ update diarization to 3.1 ([c30408f](https://github.com/melMass/comfy_mtb/commit/c30408f96d4df9c7d35545654401162090a74305)) by [@numz](https://github.com/numz)
+- ✨ add "workflow" query to /mtb/view endpoint ([eb7cf89](https://github.com/melMass/comfy_mtb/commit/eb7cf89f173b2342b04e7b61dca3d12cfaf65bdb))
+- ✨ add stretch_x and stretch_y to TransformImage ([22fce6f](https://github.com/melMass/comfy_mtb/commit/22fce6fdda135cbb1f1aad42c86aae166cba81b5))
+- ✨ add AudioDuration node ([f471497](https://github.com/melMass/comfy_mtb/commit/f47149746ac1e418cda2007c38aafbb03946ce22))
+- ✨ basic whisper nodes ([83cfc5c](https://github.com/melMass/comfy_mtb/commit/83cfc5c723d1a572af67ad14b52be4f8371a3c5f))
+- ✨ add BboxForDimensions ([9405784](https://github.com/melMass/comfy_mtb/commit/940578476438eaa6a42e0056f1b7b319ee585334))
+- ✨ improve the debug node ([cf7a9c4](https://github.com/melMass/comfy_mtb/commit/cf7a9c41e81e8dd461ab9dfa3c05bb8e2cdf2a67))
+- ✨ add BatchImageToSublist and counterpart ([00173fa](https://github.com/melMass/comfy_mtb/commit/00173fa3fbca4c5b1ff3016cc5139705ce61ec20))
+- ✨ add TensorOps ([a8cf465](https://github.com/melMass/comfy_mtb/commit/a8cf4650ff5cbd4975ef954b5829c772ee53250c))
+- ✨ live update outputs grid ([7f7a62f](https://github.com/melMass/comfy_mtb/commit/7f7a62f832c865a13b9181daee79d3cfc21581e2)) by [@christian-byrne](https://github.com/christian-byrne) in [#229](https://github.com/melMass/comfy_mtb/pull/229)
+- ✨ add SaveImage passthrough ([0eeb707](https://github.com/melMass/comfy_mtb/commit/0eeb707f34f51142def8e0ef7d351ee5028cb5e0))
+- ✨ add filtering to TransformImage ([bae26a0](https://github.com/melMass/comfy_mtb/commit/bae26a07fb02dd518c621eba28986a51c5d086bc))
+- ✨ add support for video in I/O sidebar ([c92d99a](https://github.com/melMass/comfy_mtb/commit/c92d99a8a37a64cfc285296f21452c4927a22774))
+- ✨ add an extra static input to Stack Images ([3f6d082](https://github.com/melMass/comfy_mtb/commit/3f6d08294096918d50101a19083f9134305cc8c9)) in [#222](https://github.com/melMass/comfy_mtb/pull/222)
+- ✨ add support for subdirs (i/o sidebar) ([52bd76e](https://github.com/melMass/comfy_mtb/commit/52bd76e19c8bd7e72986900e5dbfade0457ef7e0))
+- ✨ add Batch Sequence Nodes ([827c64c](https://github.com/melMass/comfy_mtb/commit/827c64c43d52ebfb8acd2e5c4491c4b66e6b8f40))
+- ✨ add support for more formats (I/O sidebar) ([8c629be](https://github.com/melMass/comfy_mtb/commit/8c629bee186b5ac991058018a788e4a836eef630))
+
+### Miscellaneous Tasks
+
+- 🧹 bump version ([d093d76](https://github.com/melMass/comfy_mtb/commit/d093d76efd87474a3ca82858147255038060ab17))
+- 🧹 small adjustments ([01107c4](https://github.com/melMass/comfy_mtb/commit/01107c45f8539ff7c579e08e2a9075d93781b9a2))
+- 🤖 update publish action workflow with permissions and version constraints ([0e48aaa](https://github.com/melMass/comfy_mtb/commit/0e48aaa3e4f1e440a5d7ab42df56b728ced03aca)) by [@robinjhuang](https://github.com/robinjhuang) in [#237](https://github.com/melMass/comfy_mtb/pull/237)
+- 🧹 basic standalone detection ([3a25526](https://github.com/melMass/comfy_mtb/commit/3a25526e818a1af8f886d2ad5c27101c4a0caa8b))
+- 🧹 rename type ([edcb3da](https://github.com/melMass/comfy_mtb/commit/edcb3da08bff66f9adcef8dcd37c3925e64d0135))
+- 🧹 update env file ([fc908ba](https://github.com/melMass/comfy_mtb/commit/fc908ba0a528523b7c1e37e34fb32f430746de0d))
+- 🧹 dev ([9a94371](https://github.com/melMass/comfy_mtb/commit/9a943714aada107bfd236e00fa1063872db7a834))
+- 🧹 apply formatting ([58ae89f](https://github.com/melMass/comfy_mtb/commit/58ae89f8e0f0f8b42825722a6aebc04da39847b1))
+
+### Refactor
+
+- 📦 add model autodownload ([147edcf](https://github.com/melMass/comfy_mtb/commit/147edcfcbc09dd27a0c787f9da568fb850c3308a))
+
+### Wip
+
+- 🚧 loop drawing ([ead4b34](https://github.com/melMass/comfy_mtb/commit/ead4b34e6dd03ea4ed309b246ef31c995325aa08))
+
+## New Contributors
+* [@poetryiii](https://github.com/poetryiii) made their first contribution in [#233](https://github.com/melMass/comfy_mtb/pull/233)
+* [@numz](https://github.com/numz) made their first contribution in [#](https://github.com/melMass/comfy_mtb/pull/)
+* [@webfiltered](https://github.com/webfiltered) made their first contribution in [#223](https://github.com/melMass/comfy_mtb/pull/223)
+## [0.2.0] - 2024-12-08
+
+### Bug Fixes
+
+- 🐛 remove mtb sidebar ([b0d52f7](https://github.com/melMass/comfy_mtb/commit/b0d52f73051368df6de2d1e10ad28ca56df72803))
+- 🐛 always enable the I/O sidebar ([ec1cb1a](https://github.com/melMass/comfy_mtb/commit/ec1cb1ac17d14670aa756dfb1ae7542397b12559))
+- 🐛 ui shifts on animation builder ([ecbb220](https://github.com/melMass/comfy_mtb/commit/ecbb220de6a05f2e506ec43f2b786be983166157))
+- 🐛 category for settings ([b6fa571](https://github.com/melMass/comfy_mtb/commit/b6fa571fd2096ace60d03cab42dba9ca37d0cb27)) in [#211](https://github.com/melMass/comfy_mtb/pull/211)
+- 🐛 new UI issues ([f272526](https://github.com/melMass/comfy_mtb/commit/f272526bfc5da95e95d42cb4c613a0b9585b2577))
+- 🐛 disable old BOOL widget (legacy) ([8596b81](https://github.com/melMass/comfy_mtb/commit/8596b8184edb484c907475a77ac1dc9e4a5c92af))
+- 🐛 pass ONNX providers explicitely ([43092e4](https://github.com/melMass/comfy_mtb/commit/43092e44a4ea17f90fcfb12372da634fe4b79557))
+- 🐛 typo in mtb_widgets error catch ([80b5a0c](https://github.com/melMass/comfy_mtb/commit/80b5a0ca7459763e7662421bccd8636976eefddd)) by [@christian-byrne](https://github.com/christian-byrne) in [#197](https://github.com/melMass/comfy_mtb/pull/197)
+- 🐛 doc widget sidebar offset in the new ui ([81b3bc1](https://github.com/melMass/comfy_mtb/commit/81b3bc1651f06ad2fa7938f810d3f406f5e7c41c))
+- 🐛 don't fallback to eval ([997d2fb](https://github.com/melMass/comfy_mtb/commit/997d2fb13af6aadf36873ea2ea3317e56f405aef))
+- 🐛 rework main utils ([c99b081](https://github.com/melMass/comfy_mtb/commit/c99b0812ab4a4183ef9298fb8a7c954bc7c858b2))
+- 🐛 MaskToImage ([821a0ef](https://github.com/melMass/comfy_mtb/commit/821a0ef42735a0a97ab82be22a4fdc67c9cfc80e))
+
+### Documentation
+
+- 📚 update wiki ([e17c6e2](https://github.com/melMass/comfy_mtb/commit/e17c6e29f5111bf5085b1fe6f764cfd1aae709f2))
+- 📚 remove link ([5bc125d](https://github.com/melMass/comfy_mtb/commit/5bc125d2f08470c8900dfd89deca721835848917))
+- 📚 clean readme ([333f646](https://github.com/melMass/comfy_mtb/commit/333f646ab1959d2c944fb046275cc93a545d557c))
+
+### Features
+
+- ✨ add h264 compression node ([e32d1e0](https://github.com/melMass/comfy_mtb/commit/e32d1e02df5e3a9351f829513f7ee3ffb2934be4))
+- ✨ add postshot nodes ([27e03fa](https://github.com/melMass/comfy_mtb/commit/27e03fa23efffda461c6975b15fe3964de476cb3))
+- ✨ improve the I/O sidebar ([cd9e614](https://github.com/melMass/comfy_mtb/commit/cd9e614b1a385d6b06eacfaad62def1d69f09808)) in [#193](https://github.com/melMass/comfy_mtb/pull/193)
+- ✨ add UpscaleBBoxBy ([74af5c6](https://github.com/melMass/comfy_mtb/commit/74af5c6499ef5dd73ce66c4c21b8c3507d69b037))
+- ✨ simplified sidebar and backend ([22f7c30](https://github.com/melMass/comfy_mtb/commit/22f7c3037345a866c9ff0b06f6689748021cee63))
+- ✨ add Interpolate Condition ([0133fb9](https://github.com/melMass/comfy_mtb/commit/0133fb93bc944d0dd7593b89b36e5b2676d9397a))
+- ✨ dump of wip things... ([cf7d305](https://github.com/melMass/comfy_mtb/commit/cf7d30507e7e449c4489e6a1ca159d3d0486bc55))
+- ✨ use the new parser for documentations ([4e593bb](https://github.com/melMass/comfy_mtb/commit/4e593bb30be561e39f1790e3514f60bb39e5a261))
+- ✨ add @mtb/markdown-parser bundles ([097ca33](https://github.com/melMass/comfy_mtb/commit/097ca33b8e7b27148e183e91712dc34d98d1a69b))
+- ✨ add VitMatte nodes ([896a025](https://github.com/melMass/comfy_mtb/commit/896a025006f9c7809c5e0776393a28f908be8950))
+- ✨ add ColorCorrectGPU ([9651a70](https://github.com/melMass/comfy_mtb/commit/9651a7034120589b059329b21688708e42772453))
+- ✨ add Swap FG/BG colors to MaskToImage ([57683c3](https://github.com/melMass/comfy_mtb/commit/57683c3c7d299a117a26526d52de4c26f2ec0f69))
+- ✨ add Extract coordinates ([f99f92e](https://github.com/melMass/comfy_mtb/commit/f99f92e8f7b2d6fac56f7f40049715910e15cfee))
+- ✨ add AudioCut ([5681b46](https://github.com/melMass/comfy_mtb/commit/5681b464adce395086712b61159b2694150b8027))
+- ✨ add AudioStack ([8d0fcee](https://github.com/melMass/comfy_mtb/commit/8d0fcee2f3decc1cbbf3b850332e6b2a022e1377))
+- ✨ add AudioSequence node ([1078fc6](https://github.com/melMass/comfy_mtb/commit/1078fc6f0fb225b52536f25ec6a9fa0456a90595))
+- ✨ add Split Bbox node ([9007a70](https://github.com/melMass/comfy_mtb/commit/9007a70aa0d6b2ead0f68f7aff8ae8e3c4f3624f))
+- ✨ update lerp example ([1a0ebd5](https://github.com/melMass/comfy_mtb/commit/1a0ebd5173687784f279a9c2184c89fb3be01dc5))
+
+### Miscellaneous Tasks
+
+- 🧹 bump minor ([50cb6f5](https://github.com/melMass/comfy_mtb/commit/50cb6f5ed6e5d9fecb9733ef3f7852b8500005e9))
+- 🧹 add worktree to gitignores ([9ccf572](https://github.com/melMass/comfy_mtb/commit/9ccf572a158caeab9bff53853e8f6fb85b76776d))
+- 🧹 remove dupe code ([e099d58](https://github.com/melMass/comfy_mtb/commit/e099d581a7627c3a66d2e3e6df3a701b0e5f31b7))
+- 🧹 update externs ([784fb01](https://github.com/melMass/comfy_mtb/commit/784fb0145b7421e2730b52237ce6a8b63b189191))
+- 🧹 add pathlibed inputs to utils ([a825504](https://github.com/melMass/comfy_mtb/commit/a825504bdd67e3461be8118119e0becc35f8af40))
+- 🧹 disable Constant ([22190cd](https://github.com/melMass/comfy_mtb/commit/22190cd25ee590595f8f19e75a9a6c539699622b))
+- 🧹 new ui is default, flag for old ui ([a976adb](https://github.com/melMass/comfy_mtb/commit/a976adbb39a13b4cd76f224ebba40c604900c862))
+- 🧹 add methods to shared ([f8829fc](https://github.com/melMass/comfy_mtb/commit/f8829fcb373e0f9bc4f0ad36c939f372349943bf))
+- 🧹 add an old_ui flag to my launcher ([dbdf276](https://github.com/melMass/comfy_mtb/commit/dbdf27664cd207dbbc69b8d635adcd59ed8d269a))
+- 🧹 move qrcode to his own file ([7d5569e](https://github.com/melMass/comfy_mtb/commit/7d5569e5c1e0f0b6ccb505a02f74640139d6aaf9))
+
+## [0.1.6] - 2024-07-03
+
+### Bug Fixes
+
+- 🐛 menu callback issue ([d64fac4](https://github.com/melMass/comfy_mtb/commit/d64fac4b74e0590acde5e3b8edd4a2f715448cf5))
+
+### Documentation
+
+- 📚 Update requirements file in INSTALL.md ([f25f6bd](https://github.com/melMass/comfy_mtb/commit/f25f6bdcd13d50f9d383065321320b0ce6a03214)) by [@elthariel](https://github.com/elthariel) in [#186](https://github.com/melMass/comfy_mtb/pull/186)
+
+### Features
+
+- ✨ add alpha channel support for faceswap/restore ([d6343e1](https://github.com/melMass/comfy_mtb/commit/d6343e1860f46947e93758f8bba03857c9326b38))
+
+### Miscellaneous Tasks
+
+- 🧹 better classname extraction ([d687497](https://github.com/melMass/comfy_mtb/commit/d687497d8041ab5d77bd31909592def6e4d0e7f6))
+- 🤖 limit release only to tags ([4eebdd8](https://github.com/melMass/comfy_mtb/commit/4eebdd8b8bff73c3db4f0248da8dac7d67cb310b))
+- 🧹 runner ([fb34671](https://github.com/melMass/comfy_mtb/commit/fb34671ee6fe80b965fe576c279ed1ff77a358f2))
+- 🤖 only publish on tag ([f1b4846](https://github.com/melMass/comfy_mtb/commit/f1b484617a917d38d9b3658d8920aa7dec672a79))
+- 🧹 small fixes ([4507842](https://github.com/melMass/comfy_mtb/commit/4507842a706141977a6a68945c36e977c358d91a))
+
+## New Contributors
+* [@elthariel](https://github.com/elthariel) made their first contribution in [#186](https://github.com/melMass/comfy_mtb/pull/186)
+## [0.1.5] - 2024-06-21
+
+### Bug Fixes
+
+- 🐛 keep the last model match instead of first ([1edc2cd](https://github.com/melMass/comfy_mtb/commit/1edc2cd10de81297e7a895009d358813e79b70ba))
+- 🐛 properly initialize the curve value ([35622e3](https://github.com/melMass/comfy_mtb/commit/35622e3a5e58103a8f5b150556b85e97e31555e1))
+- 🐛 ImageCompare improvements ([acc2d68](https://github.com/melMass/comfy_mtb/commit/acc2d687d596bf82c2075f9a24003eacf18adfe7)) by [@christian-byrne](https://github.com/christian-byrne) in [#176](https://github.com/melMass/comfy_mtb/pull/176)
+- 🐛 repetitive warning ([780c52f](https://github.com/melMass/comfy_mtb/commit/780c52f03aca3079a1b695510341486720004bec)) by [@vxkj1211](https://github.com/vxkj1211) in [#177](https://github.com/melMass/comfy_mtb/pull/177)
+- 🐛 add back was conversion node ([349a852](https://github.com/melMass/comfy_mtb/commit/349a8524c6f7fcab4a124cacb60bfbef1463cf1b))
+- 🐛 drag lag on documentation resize handle ([15330ea](https://github.com/melMass/comfy_mtb/commit/15330eab655f66214d3c25fd237679f090175c32))
+- 🐛 kwarg typo ([1571782](https://github.com/melMass/comfy_mtb/commit/1571782d012b83bce32a065e700f9a587db234d2))
+- 🐛 seed of PlotBatchFloat ([5b40302](https://github.com/melMass/comfy_mtb/commit/5b4030288d43c79859c9706a12aa0f8b7dea190f))
+- 🐛 forceInput for FLOAT <-> FLOATS converters ([5a0ef0d](https://github.com/melMass/comfy_mtb/commit/5a0ef0dadd01fd5937ed0715d829d6a456f96318))
+- 🐛 FLOAT always need options to be set ([967e72f](https://github.com/melMass/comfy_mtb/commit/967e72fc66780685f8192cb8fe13ba66b9326f63))
+- 🐛 remove doc if opened on node delete ([bee3f47](https://github.com/melMass/comfy_mtb/commit/bee3f47a14ddb92b3760098666bf75dc7d37f1e4))
+- 🐛 for documentation on HiDPI ([b11346a](https://github.com/melMass/comfy_mtb/commit/b11346aba88d9f1dac3b6b42c691979cc0978b6f))
+- 🐛 never remove input 0 of dynamic inputs ([30982fa](https://github.com/melMass/comfy_mtb/commit/30982fa48829c3fc2a6745ce5a07537a3d94b2f9))
+- 🐛 use the same fix as dynamicInputs for debug ([92b7990](https://github.com/melMass/comfy_mtb/commit/92b79906cd2ee1b4ca3ff25378d7786b5a47cb75))
+- 🐛 missing numberInput ([76f365b](https://github.com/melMass/comfy_mtb/commit/76f365b5eee165c76f3da7d2e3950786685bc08b))
+- 🐛 better curve ([da67e76](https://github.com/melMass/comfy_mtb/commit/da67e766c2f700dd9e2f51a5bafe07c612904f5d))
+- 🐛 prepend MTB_ to all classes ([b1d74ad](https://github.com/melMass/comfy_mtb/commit/b1d74adb15166e3e5eb9cf92d6148e4644bed346))
+- 🐛 dynamic connections ([652ac3f](https://github.com/melMass/comfy_mtb/commit/652ac3f3b971582b02115177fd6f7a9d3d7295df))
+- 🐛 remaining issue before merge ([100067a](https://github.com/melMass/comfy_mtb/commit/100067a645194366426f29b085bf25d0623f4fac))
+- 🐛 debug issues ([7807449](https://github.com/melMass/comfy_mtb/commit/7807449e6dcc01cfdb7f0eb818569184c8b41af2))
+- 🐛 errors when insightface's folder missing ([e838c04](https://github.com/melMass/comfy_mtb/commit/e838c04758402250fd3464d6cd6a6f872e8cef29))
+- 🐛 typo ([e40ad7a](https://github.com/melMass/comfy_mtb/commit/e40ad7a574f961ebe1f338b97214da5cbadcc529))
+- 🐛 better defaults (cont) ([1da483a](https://github.com/melMass/comfy_mtb/commit/1da483a8baa6a893f1adb05ef79b90c4412c3834))
+- 🐛 better defaults for Autopan ([5eff38b](https://github.com/melMass/comfy_mtb/commit/5eff38b387d22206d39c08e435806f9d03992feb))
+- 🐛 dynamic inputs ([9ab20a0](https://github.com/melMass/comfy_mtb/commit/9ab20a0ab50b1656ded9a84c13769fd2d547f2d2))
+- 🐛 bundle ace editor ([7c35582](https://github.com/melMass/comfy_mtb/commit/7c3558273bebc0754c802720e705232f220a0da4))
+- 🐛 image to mask ([f16d576](https://github.com/melMass/comfy_mtb/commit/f16d576f6f0e83fc2fafd2d1f29b2edeb00d3197))
+- 🐛 prepend MTB to classnames ([e56508c](https://github.com/melMass/comfy_mtb/commit/e56508c2078155f053e7f11d538a048df6a5b18b))
+- 🐛 allow smaller values in BatchTransform ([9a4b27d](https://github.com/melMass/comfy_mtb/commit/9a4b27d2e05e8ebe31f58a21db94bd3a54ed23d9))
+- 🐛 add category for virtual note+ ([eeac8c0](https://github.com/melMass/comfy_mtb/commit/eeac8c002ad1f9e461418fb66b9338e969259e58))
+- 🐛 make image feed of by default ([df0a98b](https://github.com/melMass/comfy_mtb/commit/df0a98b94a4a9388811bc8786e820ec892919c1a))
+- 🐛 support batch masks (colored image node) ([2465ffb](https://github.com/melMass/comfy_mtb/commit/2465ffb0d3b052fb78559394dbb550bba59b97a3))
+- 🐛 support pillow < 10 ([48f91b7](https://github.com/melMass/comfy_mtb/commit/48f91b74e2c7ef6d31c094eafa5332784a275a8b))
+- 🐛 image rotation bug ([54ff658](https://github.com/melMass/comfy_mtb/commit/54ff6583ded0ed4054f8e5d7fadf0b2350259dce)) by [@hongminpark](https://github.com/hongminpark) in [#154](https://github.com/melMass/comfy_mtb/pull/154)
 - 🐛 font fallback ([9fccdee](https://github.com/melMass/comfy_mtb/commit/9fccdee82d721e88c64d2292c209fec869524dd2))
 - ✨ optional inputs of colored image ([cd32f26](https://github.com/melMass/comfy_mtb/commit/cd32f26b167088d6b489e43b260c187ea5e4d223)) by [@ScottNealon](https://github.com/ScottNealon) in [#147](https://github.com/melMass/comfy_mtb/pull/147)
 - 📝 adds a way to not load the imagefeed ([501c330](https://github.com/melMass/comfy_mtb/commit/501c3301056b2851555cccd75ab3ff15b1ab8e0c))
@@ -47,6 +230,13 @@ Check the notes in the [releases](https://github.com/melMass/comfy_mtb/releases)
 
 ### Documentation
 
+- 📚 update the wiki ([fa3199b](https://github.com/melMass/comfy_mtb/commit/fa3199be2b87bf3cb7484a0fee32a8ac099adc65))
+- 📚 update wiki submodule ([49cea8d](https://github.com/melMass/comfy_mtb/commit/49cea8d94508b27781506e3b5509c65e1d84e80f))
+- 📚 add the wiki as a submodule ([5998924](https://github.com/melMass/comfy_mtb/commit/59989249260a9c579ec851c50534b58f3f02cd61))
+- 📚 missing doc ([c9836a8](https://github.com/melMass/comfy_mtb/commit/c9836a87f6823db1d53e56997417f3cbe8cc4727))
+- 📚 use flat icon ([991af4f](https://github.com/melMass/comfy_mtb/commit/991af4f45ff8c660b2c45466bb219186699170ed))
+- 📚 add banodoco channel link ([9ce34b4](https://github.com/melMass/comfy_mtb/commit/9ce34b47fd99b18db7997ccce44e6063f00b6801))
+- 📚 udpate changelog ([8221c49](https://github.com/melMass/comfy_mtb/commit/8221c49942bd87c14d5063066315a449a1fee86e))
 - 📝 add changelog ([0d817bf](https://github.com/melMass/comfy_mtb/commit/0d817bf326b4a22e2221264a414af50c3b7048b9))
 - 📄 add note+ screenshot ([90d9636](https://github.com/melMass/comfy_mtb/commit/90d96366c8b7637b55d1b4f88cb9aca217c1414b))
 - 📝 add cover image ([6b993b8](https://github.com/melMass/comfy_mtb/commit/6b993b84071bbb80ba1b8bd63576f31e35d05590))
@@ -60,6 +250,28 @@ Check the notes in the [releases](https://github.com/melMass/comfy_mtb/releases)
 
 ### Features
 
+- ✨ add ModelPruner (wip) ([43d65ae](https://github.com/melMass/comfy_mtb/commit/43d65ae68c97e077117b17b7c9d1936583f965eb))
+- ✨ Use dynamic contrast in Color Correct ([6abac2e](https://github.com/melMass/comfy_mtb/commit/6abac2e4706a3d937420213e01468bae10cc2017)) by [@christian-byrne](https://github.com/christian-byrne) in [#180](https://github.com/melMass/comfy_mtb/pull/180)
+- ✨ StackImages add support for batch mismatch ([5060c56](https://github.com/melMass/comfy_mtb/commit/5060c561353e43624ec164cb73fce7d1d422f765))
+- ✨ add BatchFloatMath ([f9d2ebf](https://github.com/melMass/comfy_mtb/commit/f9d2ebf91d09fc214fecf7501a5490b33c30aca2))
+- ✨ add FLOATS to INTS ([1b7ae27](https://github.com/melMass/comfy_mtb/commit/1b7ae27cc1907bfba3c5166ec2c61547babd2e0a))
+- ✨ debug dict ([63ee25d](https://github.com/melMass/comfy_mtb/commit/63ee25d001d4c94aa95dc8b39008f5d943f2ab45))
+- ✨ add Swap BG/FG color menu item ([1caf7c1](https://github.com/melMass/comfy_mtb/commit/1caf7c18c372651b2be7227eb77e2251d963693d))
+- ✨ BatchFloatFit the batch version of FitNumber ([ab58c36](https://github.com/melMass/comfy_mtb/commit/ab58c362124f0f4b3178534ca78cb924fb881534))
+- ✨ add FloatToFloats (the counterpart) ([78a86da](https://github.com/melMass/comfy_mtb/commit/78a86daaf71dab5be34b90b13491460854718485))
+- ✨ add some FLOATS batch nodes ([2159395](https://github.com/melMass/comfy_mtb/commit/2159395389429c5f7012e660b41fad48d376b39f))
+- ✨ poc of the doc widget idea ([fac7529](https://github.com/melMass/comfy_mtb/commit/fac7529d1f7b6fc4b3b2e7f6022ebb23ec71169d))
+- ✨ add the backend node for Constant ([dff5b22](https://github.com/melMass/comfy_mtb/commit/dff5b2201d73c1a91d4b5864e3b974e68846a011))
+- ✨ add Constant node ([cbb5dd2](https://github.com/melMass/comfy_mtb/commit/cbb5dd2cf810d5648a64eae370dba610336b99d5))
+- ✨ add FloatsToFloat ([6ebecfd](https://github.com/melMass/comfy_mtb/commit/6ebecfd8cf1dc3779384e565a65baa9dceb43660))
+- ✨ add AutoPanEquilateral ([3513937](https://github.com/melMass/comfy_mtb/commit/35139371e84d715423015e05d1b4a6c1d88b0eb5))
+- ✨ add MatchDimensions ([5db3ebe](https://github.com/melMass/comfy_mtb/commit/5db3ebedb9d38470c82544e45970775193add05c))
+- ✨ add equilateral example ([8d65556](https://github.com/melMass/comfy_mtb/commit/8d65556c37f33d1c496504db92574805916dd613))
+- ✨ enhance tiling tools ([ba73fc6](https://github.com/melMass/comfy_mtb/commit/ba73fc6af7039a4629a73cdc36a8c8736dc27c9d))
+- ✨ add FLOATS support to blur ([92c810c](https://github.com/melMass/comfy_mtb/commit/92c810c5036f7a2b3f84a3fde8c81e6a2b046b07))
+- ✨ add "tube" to Batch Shape ([f658fc3](https://github.com/melMass/comfy_mtb/commit/f658fc31e040141209384d98dfe84b766fe4ae11))
+- ✨ note+ editor themes ([133da70](https://github.com/melMass/comfy_mtb/commit/133da705c94af2dfb3d2f38c0d9c2723c72cacf7))
+- ✨ add ffmpeg gif export ([1b29aad](https://github.com/melMass/comfy_mtb/commit/1b29aad360116e631b7b4d34e98a5a631f134977)) by [@huanggou666](https://github.com/huanggou666) in [#159](https://github.com/melMass/comfy_mtb/pull/159)
 - ✨ add "To Device" ([c28181f](https://github.com/melMass/comfy_mtb/commit/c28181f1615d2e183767aa76cc2350934330e546))
 - ✨ add note+ example ([90f3bc2](https://github.com/melMass/comfy_mtb/commit/90f3bc2d953b299ea34e9e3a925f1a824b488855))
 - 💄 node+ improvements ([4b29395](https://github.com/melMass/comfy_mtb/commit/4b29395000254382882c0d1be115b2ed80cd7c99))
@@ -82,6 +294,19 @@ Check the notes in the [releases](https://github.com/melMass/comfy_mtb/releases)
 
 ### Miscellaneous Tasks
 
+- 🧹 add fields for the registry ([bb5682a](https://github.com/melMass/comfy_mtb/commit/bb5682aa6da923859db33830c2e46f24b19199a1))
+- 🧹 add pre-commit ([59612fd](https://github.com/melMass/comfy_mtb/commit/59612fd8110a888f0081433242a2b5a5f7e46da6))
+- 🧹 migrate from poetry to setuptools ([dfd17f6](https://github.com/melMass/comfy_mtb/commit/dfd17f6d783e784df7dab38d185c747b4c04d1d0))
+- 🧹 remove logs ([1070edd](https://github.com/melMass/comfy_mtb/commit/1070edd0245fb235183d5f38cd1bebf6e0405f97))
+- 🧹 add more pyproject meta ([644371e](https://github.com/melMass/comfy_mtb/commit/644371e5b5a2b8260fc5c6f699465b0bc1c81d57))
+- 🤖 move at the proper location ([f3d468c](https://github.com/melMass/comfy_mtb/commit/f3d468cfc238f13905a13a7b2225e3711129c64d))
+- 🤖 add CI to publish to ComfyUI Registry ([6cd448b](https://github.com/melMass/comfy_mtb/commit/6cd448b026956cdf3f1b81e93724b295316fbf09)) by [@haohaocreates](https://github.com/haohaocreates) in [#182](https://github.com/melMass/comfy_mtb/pull/182)
+- 🧹 add ComfyUI registry to pyproject.toml ([5951c90](https://github.com/melMass/comfy_mtb/commit/5951c90b10f9b77b2b617e83efe0112f43c8daef)) by [@haohaocreates](https://github.com/haohaocreates) in [#181](https://github.com/melMass/comfy_mtb/pull/181)
+- 🧹 update types ([96a0da9](https://github.com/melMass/comfy_mtb/commit/96a0da9dbd051d1fcf8b332c54ed2d307d8ae0dd))
+- 🧹 use a gettattr fallback ([a344cdc](https://github.com/melMass/comfy_mtb/commit/a344cdcba9823ca1fb0762795068039b1e1cf0ab))
+- 🧹 cleanup js ([64cc4e9](https://github.com/melMass/comfy_mtb/commit/64cc4e9649853023d645245bea1e1ceb11073f01))
+- 🧹 add savedatabundle js part ([edd7c3f](https://github.com/melMass/comfy_mtb/commit/edd7c3f5d075b640e9cdb067ebfe51c42ff61791))
+- 🧹 wip dynamic multitype ([71bfdd6](https://github.com/melMass/comfy_mtb/commit/71bfdd61d731ce15f9bd0bb19d65b5af208d5dcf))
 - 🧹 applied some linting ([fe49312](https://github.com/melMass/comfy_mtb/commit/fe49312cbef03c6540304448fa88aa7a88391efa))
 - 📝 header links not parsed ([514c0d2](https://github.com/melMass/comfy_mtb/commit/514c0d2eda9990435eb18258d4bbd1aa137feb3d))
 - 📝 hardcode links in changelog ([915b744](https://github.com/melMass/comfy_mtb/commit/915b7444a9db83f349d83b636304af0d276f529f))
@@ -104,9 +329,17 @@ Check the notes in the [releases](https://github.com/melMass/comfy_mtb/releases)
 
 ### Wip
 
+- 🚧 curve widget logic fixed ([e312b02](https://github.com/melMass/comfy_mtb/commit/e312b02ad2f8334e87654a20b0114837df229371))
+- 🚧 dump3 ([eedbb4b](https://github.com/melMass/comfy_mtb/commit/eedbb4bc6581bef85c746307fe9d53360ea45bcf))
+- 🚧 dump ([fa23975](https://github.com/melMass/comfy_mtb/commit/fa2397585fff4f54bcf17f0b0e0083c427b34fa8))
+- 🚧 dump ([0d0fb8e](https://github.com/melMass/comfy_mtb/commit/0d0fb8e13a5da54a44a96a04607f7a349f8fdb03))
 - 🚧 add text template node ([af2175a](https://github.com/melMass/comfy_mtb/commit/af2175a1fc0c2fb29ef3493f242fe45ec6fcabac))
 
 ## New Contributors
+* [@haohaocreates](https://github.com/haohaocreates) made their first contribution in [#182](https://github.com/melMass/comfy_mtb/pull/182)
+* [@vxkj1211](https://github.com/vxkj1211) made their first contribution in [#177](https://github.com/melMass/comfy_mtb/pull/177)
+* [@huanggou666](https://github.com/huanggou666) made their first contribution in [#159](https://github.com/melMass/comfy_mtb/pull/159)
+* [@hongminpark](https://github.com/hongminpark) made their first contribution in [#154](https://github.com/melMass/comfy_mtb/pull/154)
 * [@ScottNealon](https://github.com/ScottNealon) made their first contribution in [#147](https://github.com/melMass/comfy_mtb/pull/147)
 * [@Yurchikian](https://github.com/Yurchikian) made their first contribution in [#124](https://github.com/melMass/comfy_mtb/pull/124)
 * [@M1kep](https://github.com/M1kep) made their first contribution in [#91](https://github.com/melMass/comfy_mtb/pull/91)
@@ -393,7 +626,10 @@ Check the notes in the [releases](https://github.com/melMass/comfy_mtb/releases)
 
 - 🚀 add gh action ([572b4d5](https://github.com/melMass/comfy_mtb/commit/572b4d52bce1398660d4d7ca0c5c48c11e0128e3)) in [#4](https://github.com/melMass/comfy_mtb/pull/4)
 
-[main]: https://github.com/melMass/comfy_mtb/compare/v0.1.4..main
+[main]: https://github.com/melMass/comfy_mtb/compare/v0.2.0..main
+[0.2.0]: https://github.com/melMass/comfy_mtb/compare/v0.1.6..v0.2.0
+[0.1.6]: https://github.com/melMass/comfy_mtb/compare/v0.1.5..v0.1.6
+[0.1.5]: https://github.com/melMass/comfy_mtb/compare/v0.1.4..v0.1.5
 [0.1.4]: https://github.com/melMass/comfy_mtb/compare/v0.1.3..v0.1.4
 [0.1.3]: https://github.com/melMass/comfy_mtb/compare/v0.1.2..v0.1.3
 [0.1.2]: https://github.com/melMass/comfy_mtb/compare/v0.1.1..v0.1.2

CODE_OF_CONDUCT.md

@@ -0,0 +1,52 @@
+# Code of Conduct
+
+## Our Commitment
+
+We are committed to creating a welcoming and inclusive community for everyone. We believe that a diverse and respectful community is essential for fostering creativity and innovation. We expect all members of our community to adhere to this Code of Conduct.
+
+## Our Expectations
+
+This Code of Conduct applies to all interactions within the mtb community, including:
+
+*   Public communication channels (e.g., GitHub issues, pull requests, discussions, social media)
+*   Private communication channels (e.g., direct messages, email)
+*   In-person events (if any)
+
+We expect all members to:
+
+*   **Be respectful and considerate:** Treat others with kindness and empathy.
+*   **Be inclusive:** Welcome and respect people of all backgrounds, identities, and experiences.
+*   **Be constructive:** Focus on providing helpful and positive feedback.
+*   **Be mindful of your language:** Avoid using offensive, discriminatory, or harassing language.
+*   **Respect privacy:** Do not share personal information without consent.
+
+## Unacceptable Behavior
+
+The following behaviors are not tolerated:
+
+*   Offensive, discriminatory, or harassing language or conduct
+*   Personal attacks or insults
+*   Spamming or trolling
+*   Sharing of malicious or inappropriate content
+*   Disrupting the community or hindering collaboration
+*   Violating the privacy of others
+
+## Reporting Violations
+
+If you experience or witness a violation of this Code of Conduct, please report it to @melmass.  All reports will be treated confidentially and investigated promptly.
+
+## Enforcement
+
+Violations of this Code of Conduct may result in the following actions:
+
+*   Warning
+*   Removal from the community
+*   Ban from the community
+
+## License
+[![Contributor Covenant](https://img.shields.io/badge/Contributor%20Covenant-2.1-4baaaa.svg)](code_of_conduct.md)
+
+## Contact
+
+If you have any questions or concerns about this Code of Conduct, please contact @melmass.
+

CONTRIBUTING.md

@@ -0,0 +1,62 @@
+# Contributing to mtb
+
+Thank you for your interest in contributing to mtb!  We appreciate your help in making this project better.  This document outlines how you can contribute to the project.
+
+## Project Overview
+
+This project is a collection of custom nodes for ComfyUI, tailored specifically for animation workflows. It aims to provide a streamlined and user-friendly experience for creating animations within the ComfyUI environment.
+
+## Ways to Contribute
+
+We welcome all kinds of contributions! Here's how you can get involved:
+
+*   **Bug Reports:**  If you encounter any issues, please create a new issue on GitHub.  Please include clear steps to reproduce the bug, along with any relevant error messages, workflows or screenshots.
+*   **Feature Requests:**  Have an idea for a new node or feature?  Create a new issue to discuss it!  Please describe the feature in detail, and explain how it would benefit the project.
+*   **Documentation Improvements:**  Help us improve the documentation by fixing errors, adding examples, or clarifying explanations.
+*   **Code Contributions:**  We welcome contributions to the codebase!  Please see the "Development Setup" and "File Structure" sections below for more information.
+*   **Testing:**  Help us ensure the stability and reliability of the project by testing new features and bug fixes.
+*   **Refactoring:**  Help us improve the codebase by refactoring existing code to improve readability, maintainability, and performance.
+
+## Development Setup
+
+```sh
+git clone --recursive https://github.com/melmass/comfy_mtb
+```
+
+## File Structure
+
+Understanding the project structure is crucial for making effective contributions.
+
+*   **`./nodes/*.py`:** This directory contains the definitions for all custom nodes. Nodes are automatically registered when a file defines an array named `__nodes__` containing the node classes.  Make sure your node follows the ComfyUI node definition structure.
+*   **`./web/*.js`:** This directory contains all the frontend JavaScript code for the extension's user interface.
+*   **`./wiki`:** This directory is a Git submodule that contains the project's Wiki documentation, written in Markdown.  Node documentation should be created or updated in the corresponding Markdown files within this submodule. This is then referenced by the UI for in-GUI help
+
+## Coding Style
+
+We use **Ruff** for code formatting to ensure consistency.  Please run Ruff on your code before submitting a pull request.  No specific configuration is required, so the default Ruff settings will be used.
+
+## Contribution Workflow
+
+1.  **Create a Branch:** Create a new branch for your feature or fix.  Use a descriptive branch name (e.g., `feature/new-node`, `fix/bug-in-ui`). **Do not fork the main branch directly.**
+2.  **Make Changes:** Implement your changes in your branch.
+3.  **Run Tests:** (Add instructions on how to run tests if available.)
+4.  **Format Code:** Run Ruff on your code to ensure it is properly formatted.
+5.  **Create a Pull Request:** Submit a pull request to the `main` branch.  Please provide a clear and concise description of your changes.
+
+## Code of Conduct
+
+We are committed to creating a welcoming and inclusive community.  We expect all contributors to adhere to a respectful and professional code of conduct. (Consider adding a link to a CODE_OF_CONDUCT.md file or a standard code of conduct.)
+
+## Tools and Libraries
+
+*   **Python:** The primary programming language for this project.
+*   **ComfyUI:** The underlying framework for the custom nodes.
+
+## Current Focus
+
+We are currently focused on a major refactor to clean up the project's codebase. Contributions related to this effort are particularly welcome!
+
+## Thank You!
+
+Thank you for considering contributing to mtb! Your contributions are greatly appreciated.  We look forward to reviewing your pull requests!
+

__init__.py

@@ -3,11 +3,11 @@
 # File: __init__.py
 # Project: comfy_mtb
 # Author: Mel Massadian
-# Copyright (c) 2023 Mel Massadian
+# Copyright (c) 2023-2025 Mel Massadian
 #
 ###
 
-__version__ = "0.2.0"
+__version__ = "0.5.0"
 
 import os
 
@@ -31,9 +31,19 @@ from importlib import reload
 from pathlib import Path
 
 from aiohttp import web
-from server import PromptServer
+
+IN_COMFY = False
+
+try:
+    from server import PromptServer
+
+    IN_COMFY = True
+except ModuleNotFoundError:
+    IN_COMFY = False
+
 
 from .endpoint import endlog
+from .install import get_node_dependencies
 from .log import blue_text, cyan_text, get_label, get_summary, log
 from .utils import comfy_dir, here
 
@@ -230,7 +240,7 @@ if failed:
 # - ENDPOINT
 
 
-if hasattr(PromptServer, "instance"):
+if IN_COMFY and hasattr(PromptServer, "instance"):
     img_cache = None
     prompt_cache = None
 
@@ -240,16 +250,7 @@ if hasattr(PromptServer, "instance"):
         img_cache = TTLCache(maxsize=100, ttl=5)  # 1 min TTL
         prompt_cache = TTLCache(maxsize=100, ttl=5)  # 1 min TTL
 
-    restore_deps = ["basicsr"]
-    onnx_deps = ["onnxruntime"]
-    swap_deps = ["insightface"] + onnx_deps
-    node_dependency_mapping = {
-        "QrCode": ["qrcode"],
-        "DeepBump": onnx_deps,
-        "FaceSwap": swap_deps,
-        "LoadFaceSwapModel": swap_deps,
-        "LoadFaceAnalysisModel": restore_deps,
-    }
+    node_dependency_mapping = get_node_dependencies()
 
     PromptServer.instance.app.router.add_static(
         "/mtb-assets/", path=(here / "html").as_posix()
@@ -472,6 +473,26 @@ if hasattr(PromptServer, "instance"):
         if not os.path.isfile(file):
             return web.Response(status=404)
 
+        ret_workflow = request.rel_url.query.get("workflow")
+
+        if ret_workflow:
+            image = Image.open(file)
+            prompt = image.info.get("prompt", "")
+            workflow = image.info.get("workflow", "")
+
+            if workflow:
+                workflow = json.loads(workflow)
+
+            if prompt:
+                prompt = json.loads(prompt)
+
+            return web.json_response(
+                {
+                    "prompt": prompt,
+                    "workflow": workflow,
+                }
+            )
+
         preview_info = None
         if "preview" in request.rel_url.query:
             preview_params = request.rel_url.query["preview"].split(";")

endpoint.py

@@ -1,17 +1,21 @@
 import csv
 import secrets
 import sys
+import urllib.parse
 from pathlib import Path
-from typing import Any
+from typing import Any, Literal
 
+import folder_paths
 from aiohttp import web
 
+from .install import get_node_dependencies
 from .log import mklog
 from .utils import (
+    SortMode,
     backup_file,
+    build_glob_patterns,
+    glob_multiple,
     import_install,
-    input_dir,
-    output_dir,
     reqs_map,
     run_command,
     styles_dir,
@@ -23,7 +27,7 @@ endlog = mklog("mtb endpoint")
 import_install("requirements")
 
 
-def ACTIONS_installDependency(dependency_names=None):
+def ACTIONS_installDependency(dependency_names: list[str] | None = None):
     if dependency_names is None:
         # return web.Response(text="No dependency name provided", status=400)
         return {"error": "No dependency name provided"}
@@ -31,6 +35,14 @@ def ACTIONS_installDependency(dependency_names=None):
     endlog.debug(f"Received Install Dependency request for {dependency_names}")
     # reqs = []
     resolved_names = [reqs_map.get(name, name) for name in dependency_names]
+    allowed_deps = list(
+        {d for dep in get_node_dependencies().values() for d in dep}
+    )
+    for dep in dependency_names:
+        if dep not in allowed_deps:
+            return {
+                "error": f"Unknown dependency: {dep}, you can only use this endpoint to install {allowed_deps}"
+            }
     try:
         run_command(
             [Path(sys.executable), "-m", "pip", "install"] + resolved_names
@@ -55,59 +67,103 @@ def ACTIONS_installDependency(dependency_names=None):
     #             break
 
 
+def ACTIONS_getUserImageFolders():
+    input_dir = Path(folder_paths.get_input_directory())
+    output_dir = Path(folder_paths.get_output_directory())
+
+    input_subdirs = [x.name for x in input_dir.iterdir() if x.is_dir()]
+    output_subdirs = [x.name for x in output_dir.iterdir() if x.is_dir()]
+
+    return {"input": input_subdirs, "output": output_subdirs}
+
+
+def ACTIONS_getUserVideos(
+    size=256, count=200, offset=0, sort: str | None = None
+):
+    count = count or 1000
+    video_extensions = ["webm", "mp4", "mkv", "mov"]
+    entries = {}
+    patterns = build_glob_patterns(video_extensions)
+    input_dir = Path(folder_paths.get_input_directory())
+    entries = glob_multiple(input_dir, patterns)
+
+    sort_mode = SortMode.from_str(sort)
+
+    if sort_mode:
+        sort_key = {
+            SortMode.MODIFIED: lambda x: x.stat().st_mtime,
+            SortMode.MODIFIED_REVERSE: lambda x: x.stat().st_mtime,
+            SortMode.NAME: lambda x: x.name,
+            SortMode.NAME_REVERSE: lambda x: x.name,
+        }.get(sort_mode)
+        if sort_key:
+            reverse = sort_mode in (SortMode.MODIFIED, SortMode.NAME_REVERSE)
+            entries = sorted(entries, key=sort_key, reverse=reverse)
+
+    videos = {
+        video.name: (
+            f"/view?force_rate=0&frame_load_cap=0&skip_first_frames=0&select_every_nth=1&filename={urllib.parse.quote_plus(video.name)}&type=input&format=video&force_size={size}x?"
+        )
+        for i, video in enumerate(entries)
+        if offset <= i < offset + count
+    }
+    return videos
+
+
 def ACTIONS_getUserImages(
-    mode: str,
-    count=200,
+    mode: Literal["input", "output"],
+    count=1000,
     offset=0,
     sort: str | None = None,
     include_subfolders: bool = False,
+    subfolder=None,
 ):
     # enabled = "MTB_EXPOSE" in os.environ
     # if not enabled:
     #     return {"error": "Session not authorized to getInputs"}
 
     imgs = {}
-    entry_dir = input_dir if mode == "input" else output_dir
-    pattern = "**/*.png" if include_subfolders else "*.png"
+    count = count or 1000
 
-    entry_gen = entry_dir.glob(pattern)
+    input_dir = Path(folder_paths.get_input_directory())
+    output_dir = Path(folder_paths.get_output_directory())
+
+    entry_dir = input_dir if mode == "input" else output_dir
+    if subfolder:
+        entry_dir = entry_dir / subfolder
+
+        if not entry_dir.exists():
+            return {
+                "error": f"Subfolder {entry_dir.name} doesn't exists in {entry_dir.parent.as_posix()}"
+            }
+    supported = ["png", "jpg", "jpeg", "webp", "gif"]
 
     entries = {}
+    patterns = build_glob_patterns(supported, recursive=include_subfolders)
+    entries = glob_multiple(entry_dir, patterns)
 
-    if sort:
-        sort = sort.lower()
-        if sort == "none":
-            entries = entry_gen
-        elif sort == "modified":
-            entries = sorted(
-                entry_gen, key=lambda x: x.stat().st_mtime, reverse=True
-            )
-        elif sort == "modified-reverse":
-            entries = sorted(entry_gen, key=lambda x: x.stat().st_mtime)
-        elif sort == "name":
-            entries = sorted(entry_gen, key=lambda x: x.name)
-        elif sort == "name-reverse":
-            entries = sorted(entry_gen, key=lambda x: x.name, reverse=True)
-        else:
-            endlog.warning(f"Sort mode {sort} not supported")
-            entries = entry_gen
-    else:
-        entries = entry_gen
+    sort_mode = SortMode.from_str(sort)
 
-    for i, img in enumerate(entries):
-        if i < offset:
-            continue
+    if sort_mode:
+        sort_key = {
+            SortMode.MODIFIED: lambda x: x.stat().st_mtime,
+            SortMode.MODIFIED_REVERSE: lambda x: x.stat().st_mtime,
+            SortMode.NAME: lambda x: x.name,
+            SortMode.NAME_REVERSE: lambda x: x.name,
+        }.get(sort_mode)
+        if sort_key:
+            reverse = sort_mode in (SortMode.MODIFIED, SortMode.NAME_REVERSE)
+            entries = sorted(entries, key=sort_key, reverse=reverse)
 
-        subfolder = (
-            img.parent.relative_to(entry_dir) if include_subfolders else ""
-        )
-        imgs[img.stem] = (
-            f"/mtb/view?filename={img.name}&width=512&type={mode}&subfolder="
-            f"{subfolder}"
+    imgs = {
+        img.name: (
+            f"/mtb/view?filename={img.name}&width=512&type={mode}&subfolder={subfolder or ''}"
+            f"{img.parent.relative_to(entry_dir) if include_subfolders else ''}"
             f"&preview=&rand={secrets.randbelow(424242)}"
         )
-        if i >= count + offset - 1:
-            break
+        for i, img in enumerate(entries)
+        if offset <= i < offset + count
+    }
     return imgs
 
 
@@ -347,7 +403,7 @@ def render_table(table_dict: dict[str, Any], sort=True, title=None):
             if "dependencies" in item:
                 table_rows += f"<tr><td>{name}</td><td>"
                 table_rows += (
-                    f"{dependencies_button(name,item['dependencies'])}"
+                    f"{dependencies_button(name, item['dependencies'])}"
                 )
 
                 table_rows += "</td></tr>"

env.nu

@@ -1,5 +1,7 @@
 # NOTE: This file is only use for development you can ignore it
 
+use private/log.nu
+
 def get_root [--clean] {
     if $clean {
         $env.COMFY_CLEAN_ROOT
@@ -21,12 +23,56 @@ export def "comfy dev-web" [] {
     npm run dev
 }
 
+export def "daily run"  [] {
+  let res = (comfy update --rebase)
+  comfy update --clean
+  comfy update_extensions
+
+  daily commit $res.from $res.to
+}
+
+def short-date [] {
+  format date "%Y-%m-%d"
+}
+
+# was daily run today?
+export def "daily was-run" [] {
+
+  let daily = ($env.COMFY_MTB | path join daily.nuon)
+
+  if ($daily | path exists) {
+    let last = (open $daily | sort-by date | get date  | last | short-date)
+    let today = (date now | short-date)
+    return ($last == $today)
+  }
+  return false
+}
+
+export def "daily commit"  [from:string, to:string] {
+  let daily = ($env.COMFY_MTB | path join daily.nuon)
+  let commit = [{date: (date now) from:$from to:$to}]
+
+  let dailies = (if ($daily | path exists) {
+    open $daily | append $commit
+  } else {
+      $commit
+  })
+
+  $dailies | save -f $daily
+  log success "Commited daily check"
+}
 
 # start the comfy server
-export def "comfy start" [--clean,--old-ui, --listen] {
-
+export def "comfy start" [--clean,--old-ui, --listen, --skip-daily(-s)] {
+  if (not (daily was-run)) and not $skip_daily {
+      log info "Running daily checks"
+      daily run
+    }
     let root = get_root --clean=($clean)
     cd $root
+
+    log info "Running Server"
+
     MTB_DEBUG=true python main.py --port 3000 ...(if $old_ui { ["--front-end-version", "Comfy-Org/ComfyUI_legacy_frontend@latest"]} else {[ --front-end-version Comfy-Org/ComfyUI_frontend@latest]}) --preview-method auto ...(if $listen {["--listen"]} else {[]})
 }
 
@@ -35,75 +81,93 @@ export def "comfy update" [
     --clean # ??
     --rebase # Rebase instead of merge
 ] {
-    let root = get_root --clean=($clean)
-    let models = $"($root)/models"
-    let inputs = $"($root)/input"
-    cd $root
-    let branch_name = (git rev-parse --abbrev-ref HEAD | str trim)
-    print $"(ansi yellow_italic)Backing up and removing models symlinks(ansi reset)"
+  let root = get_root --clean=$clean
 
-    if not $clean {
-        cd $models
-        # find all symlinks
-        let links = (ls -la |
-            where not ($it.target | is-empty) |
-            select name target |
-            sort-by name)
+  let models = $"($root)/models"
+  let inputs = $"($root)/input"
+
+  cd $root
+
+  let branch_name = (git rev-parse --abbrev-ref HEAD | str trim)
+  let current_commit = (git rev-parse HEAD | str trim)
+
+  log info "Backing up and removing models symlinks"
+
+  # preparing root for pull
+  if not $clean {
+    git checkout pyproject.toml
+    cd $models
+    # find and store all symlinks
+    let links = (ls -la |
+      where not ($it.target | is-empty) |
+      select name target |
+      sort-by name)
 
 
-            if not ($links | is-empty) {
-              $links | save -f links.nuon
-              # remove them
-              open links.nuon | each {|p| rm $p.name }
-            }
-    } else {
-        rm $models
-        rm $inputs
+    if not ($links | is-empty) {
+      $links | save -f links.nuon
+      # remove them
+      open links.nuon | each {|p| rm $p.name }
     }
+  } else {
+    # just remove symlinks
+    rm $models
+    rm $inputs
+  }
 
-    cd $root
+  cd $root
 
-    print $"(ansi yellow_italic)Checking out to master(ansi reset)"
-    git checkout master
+  log info $"Checking out to master"
+  git checkout master
 
-    print $"(ansi yellow_italic)Fetching and pulling remote updates(ansi reset)"
-    if ($clean) {
-        git fetch local master
-        git pull local master
-    } else {
-        git fetch
-        git pull
-    }
+  log info "Fetching and pulling remote updates"
+  if ($clean) {
+    # from the local base repo master
+    git fetch local master # $branch_name # master
+    git pull local master # $branch_name # master
+  } else {
+    git fetch
+    git pull
+  }
 
+  let new_commit = (git rev-parse HEAD | str trim)
 
-    print $"(ansi yellow_italic)Back to our branch \(($branch_name)\)(ansi reset)"
-    git checkout -
+  log info $"Back to our branch \(($branch_name)\)"
+  git checkout -
 
+  if $current_commit == $new_commit {
+    log warn "No changes upstream"
+  } else {
     if $rebase {
-        print $"(ansi yellow_italic)Rebasing changes(ansi reset)"
-        git rebase master
+      log info "Rebasing changes"
+      git rebase master
 
     } else {
-        print $"(ansi yellow_italic)Merging changes(ansi reset)"
-        git merge master
+      log info "Merging changes"
+      git merge master
     }
+  }
 
-    print $"(ansi yellow_italic)Linking back the models(ansi reset)"
+  log info "Linking back the models"
 
-    if not $clean {
-        cd $models
-        # resymlink them
-        open links.nuon | each {|p| link -a $p.target $p.name }
-    } else {
-        let master = (get_root)
-        link ($master | path join models) $models
-        link ($master | path join input) $inputs
-    }
+  if not $clean {
+    rm pyproject.toml
+    cp pyproject-mel.toml pyproject.toml
+    cd $models
 
-    let commit_count = (git rev-list --count $branch_name $"^origin/($branch_name)")
+    # resymlink them
+    open links.nuon | each {|p| link -a $p.target $p.name }
+  } else {
+    let master = (get_root)
+    link ($master | path join models) $models
+    link ($master | path join input) $inputs
+  }
 
+  let commit_count = (git rev-list --count $branch_name $"^origin/($branch_name)")
 
-    print $"(ansi green_bold)Update successful \(($commit_count) new commits\)(ansi reset)"
+  log success $"Update successful \(($commit_count) new commits\)"
+
+  return {from:$current_commit to:$new_commit}
 
 
 }
@@ -118,11 +182,11 @@ export def "comfy toggle_extensions" [--clean] {
         return
     }
 
-    print $choices
+    log info "Choices" $choices
 
     let filtered = $choices | wrap name | upsert enabled {|p| not ($p.name | str ends-with ".disabled")}
 
-    print $filtered
+    log info "Filtered" $filtered
     $filtered | each {|f|
         let new_name = ($f.name | str replace ".disabled" "")
 
@@ -131,7 +195,7 @@ export def "comfy toggle_extensions" [--clean] {
         } else {
             $new_name
         }
-        print $"Moving ($f.name) to ($new_name)"
+        log info $"Moving ($f.name) to ($new_name)"
         mv $f.name $new_name
     }
 }
@@ -151,6 +215,7 @@ def --env path-add [pth] {
 
 
 export-env {
+  $env.PYTHONUTF8 = 1
   $env.COMFY_MTB = ("." | path expand)
   # $env.CUDA_ROOT =  'C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1\'
 
@@ -160,6 +225,12 @@ export-env {
   $env.COMFY_CLEAN_ROOT =  ($env.COMFY_ROOT | path dirname | path join ComfyClean)
 
   path-add 'C:/Portable/TensorRT-8.6.0.12/lib'
+
+  if $nu.os-info.family == 'windows' {
+    path-add 'G:\BIN\TensorRT-10.7.0.23\lib'
+    path-add 'G:\BIN\cudnn-windows-x86_64-9.6.0.74_cuda12-archive\bin'
+  }
+
   path-add ($env.CUDA_ROOT | path join bin)
   overlay use ../../.venv/Scripts/activate.nu
 }

install.py

@@ -43,10 +43,28 @@ pip_map = {
     "tb-nightly": "tensorboard",
     "protobuf": "google.protobuf",
     "qrcode[pil]": "qrcode",
-    "requirements-parser": "requirements"
+    "requirements-parser": "requirements",
     # Add more mappings as needed
 }
 
+
+def get_node_dependencies():
+    restore_deps = ["basicsr"]
+    onnx_deps = ["onnxruntime"]
+    swap_deps = ["insightface"] + onnx_deps
+    quant_deps = ["bitsandbytes"]
+    io_deps = ["av"]
+    return {
+        "QrCode": ["qrcode"],
+        "DeepBump": onnx_deps,
+        "FaceSwap": swap_deps,
+        "LoadFaceSwapModel": swap_deps,
+        "LoadFaceAnalysisModel": restore_deps,
+        "Quantize": quant_deps,
+        "SaveGif": io_deps,
+    }
+
+
 # endregion
 
 # region ansi
@@ -124,12 +142,12 @@ def print_formatted(text, *formats, color=None, background=None, **kwargs):
     header = "[mtb install] "
 
     # Handle console encoding for Unicode characters (utf-8)
-    encoded_header = header.encode(sys.stdout.encoding, errors="replace").decode(
-        sys.stdout.encoding
-    )
-    encoded_text = formatted_text.encode(sys.stdout.encoding, errors="replace").decode(
-        sys.stdout.encoding
-    )
+    encoded_header = header.encode(
+        sys.stdout.encoding, errors="replace"
+    ).decode(sys.stdout.encoding)
+    encoded_text = formatted_text.encode(
+        sys.stdout.encoding, errors="replace"
+    ).decode(sys.stdout.encoding)
 
     print(
         " " * len(encoded_header)
@@ -163,7 +181,9 @@ def run_command(cmd, ignored_lines_start=None):
     try:
         _run_command(shell_cmd, ignored_lines_start)
     except subprocess.CalledProcessError as e:
-        print(f"Command failed with return code: {e.returncode}", file=sys.stderr)
+        print(
+            f"Command failed with return code: {e.returncode}", file=sys.stderr
+        )
         print(e.stderr.strip(), file=sys.stderr)
 
     except KeyboardInterrupt:
@@ -238,7 +258,7 @@ def suppress_std():
 def get_local_version():
     init_file = os.path.join(os.path.dirname(__file__), "__init__.py")
     if os.path.isfile(init_file):
-        with open(init_file, "r") as f:
+        with open(init_file) as f:
             tree = ast.parse(f.read())
             for node in ast.walk(tree):
                 if isinstance(node, ast.Assign):
@@ -256,13 +276,16 @@ def download_file(url, file_name):
     with requests.get(url, stream=True) as response:
         response.raise_for_status()
         total_size = int(response.headers.get("content-length", 0))
-        with open(file_name, "wb") as file, tqdm(
-            desc=file_name.stem,
-            total=total_size,
-            unit="B",
-            unit_scale=True,
-            unit_divisor=1024,
-        ) as progress_bar:
+        with (
+            open(file_name, "wb") as file,
+            tqdm(
+                desc=file_name.stem,
+                total=total_size,
+                unit="B",
+                unit_scale=True,
+                unit_divisor=1024,
+            ) as progress_bar,
+        ):
             for chunk in response.iter_content(chunk_size=8192):
                 file.write(chunk)
                 progress_bar.update(len(chunk))
@@ -302,7 +325,9 @@ def import_or_install(requirement, dry=False):
     pip_install_name = pip_name + pip_spec
 
     if not installed:
-        print_formatted(f"Installing package {pip_name}...", "italic", color="yellow")
+        print_formatted(
+            f"Installing package {pip_name}...", "italic", color="yellow"
+        )
         if dry:
             print_formatted(
                 f"Dry-run: Package {pip_install_name} would be installed (import name: '{import_name}').",
@@ -310,7 +335,9 @@ def import_or_install(requirement, dry=False):
             )
         else:
             try:
-                run_command([executable, "-m", "pip", "install", pip_install_name])
+                run_command(
+                    [executable, "-m", "pip", "install", pip_install_name]
+                )
                 print_formatted(
                     f"Package {pip_install_name} installed successfully using pip package name  (import name: '{import_name}')",
                     "bold",
@@ -326,13 +353,9 @@ def import_or_install(requirement, dry=False):
 
 def get_github_assets(tag=None):
     if tag:
-        tag_url = (
-            f"https://api.github.com/repos/{repo_owner}/{repo_name}/releases/tags/{tag}"
-        )
+        tag_url = f"https://api.github.com/repos/{repo_owner}/{repo_name}/releases/tags/{tag}"
     else:
-        tag_url = (
-            f"https://api.github.com/repos/{repo_owner}/{repo_name}/releases/latest"
-        )
+        tag_url = f"https://api.github.com/repos/{repo_owner}/{repo_name}/releases/latest"
     response = requests.get(tag_url)
     if response.status_code == 404:
         # print_formatted(
@@ -361,7 +384,9 @@ except ImportError:
 def main():
     if len(sys.argv) == 1:
         print_formatted(
-            "mtb doesn't need an install script anymore.", "italic", color="yellow"
+            "mtb doesn't need an install script anymore.",
+            "italic",
+            color="yellow",
         )
         return
     if all(arg not in ("-p", "--path") for arg in sys.argv):
@@ -384,7 +409,7 @@ def main():
 
     args = parser.parse_args()
 
-    print_formatted(f"Detected environment: {apply_color(mode,'cyan')}")
+    print_formatted(f"Detected environment: {apply_color(mode, 'cyan')}")
 
     if args.path:
         clone_dir = Path(args.path)
@@ -397,8 +422,12 @@ def main():
         else:
             repo_dir = clone_dir / repo_name
             if not repo_dir.exists():
-                print_formatted(f"Cloning to {repo_dir}...", "italic", color="yellow")
-                run_command(["git", "clone", "--recursive", repo_url, repo_dir])
+                print_formatted(
+                    f"Cloning to {repo_dir}...", "italic", color="yellow"
+                )
+                run_command(
+                    ["git", "clone", "--recursive", repo_url, repo_dir]
+                )
             else:
                 print_formatted(
                     f"Directory {repo_dir} already exists, we will update it..."
@@ -409,7 +438,14 @@ def main():
 
     print_formatted("Checking environment...", "italic", color="yellow")
     missing_deps = []
-    install_cmd = [executable, "-m", "pip", "install", "-r", "requirements.txt"]
+    install_cmd = [
+        executable,
+        "-m",
+        "pip",
+        "install",
+        "-r",
+        "requirements.txt",
+    ]
     run_command(install_cmd)
 
     print_formatted(

nodes/audio.py

@@ -1,17 +1,42 @@
-from typing import TypedDict
+from typing import Any, TypedDict
 
 import torch
 import torchaudio
+from comfy.model_management import get_torch_device
+from huggingface_hub import snapshot_download
+from transformers import (
+    WhisperForConditionalGeneration,
+    WhisperProcessor,
+)
+
+# from transformers import (
+#     AutoFeatureExtractor,
+#     WhisperForConditionalGeneration,
+#     WhisperModel,
+#     WhisperProcessor,
+# )
+from ..log import log
+from ..utils import get_model_path
+
+WHISPER_SAMPLE_RATE = 16000
 
 
-class AudioDict(TypedDict):
+class AudioTensor(TypedDict):
     """Comfy's representation of AUDIO data."""
 
     sample_rate: int
     waveform: torch.Tensor
 
 
-AudioData = AudioDict | list[AudioDict]
+class WhisperData(TypedDict):
+    """Whisper transcription data with timestamps and speaker info."""
+
+    text: str
+    chunks: list[dict[str, Any]]
+    language: str
+
+
+AudioData = AudioTensor | list[AudioTensor]
 
 
 class MtbAudio:
@@ -28,10 +53,14 @@ class MtbAudio:
             return audios["waveform"].shape[1] == 2
 
     @staticmethod
-    def resample(audio: AudioDict, common_sample_rate: int) -> AudioDict:
-        if audio["sample_rate"] != common_sample_rate:
+    def resample(audio: AudioTensor, common_sample_rate: int) -> AudioTensor:
+        current_rate = audio["sample_rate"]
+        if current_rate != common_sample_rate:
+            log.debug(
+                f"Resampling audio from {current_rate} to {common_sample_rate}"
+            )
             resampler = torchaudio.transforms.Resample(
-                orig_freq=audio["sample_rate"], new_freq=common_sample_rate
+                orig_freq=current_rate, new_freq=common_sample_rate
             )
             return {
                 "sample_rate": common_sample_rate,
@@ -41,7 +70,7 @@ class MtbAudio:
             return audio
 
     @staticmethod
-    def to_stereo(audio: AudioDict) -> AudioDict:
+    def to_stereo(audio: AudioTensor) -> AudioTensor:
         if audio["waveform"].shape[1] == 1:
             return {
                 "sample_rate": audio["sample_rate"],
@@ -54,8 +83,8 @@ class MtbAudio:
 
     @classmethod
     def preprocess_audios(
-        cls, audios: list[AudioDict]
-    ) -> tuple[list[AudioDict], bool, int]:
+        cls, audios: list[AudioTensor]
+    ) -> tuple[list[AudioTensor], bool, int]:
         max_sample_rate = max([audio["sample_rate"] for audio in audios])
 
         resampled_audios = [
@@ -69,6 +98,383 @@ class MtbAudio:
         return (audios, is_stereo, max_sample_rate)
 
 
+class WhisperPipeline(TypedDict):
+    """Whisper model pipeline."""
+
+    processor: WhisperProcessor
+    model: WhisperForConditionalGeneration
+
+
+class MTB_LoadWhisper:
+    """Load Whisper model and processor."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "model_size": (
+                    [
+                        "tiny",
+                        "small",
+                        "medium",
+                        "medium.en",
+                        "base",
+                        "large",
+                        "large-v2",
+                        "large-v3",
+                        "large-v3-turbo",
+                    ],
+                    {"default": "tiny"},
+                ),
+            },
+            "optional": {
+                "download_missing": (
+                    "BOOLEAN",
+                    {
+                        "default": False,
+                        "tooltip": (
+                            "Download missing models if missing,"
+                            "otherwise they must be in ComfyUI/models/whisper"
+                        ),
+                    },
+                ),
+            },
+        }
+
+    RETURN_TYPES = ("WHISPER_PIPELINE",)
+    RETURN_NAMES = ("pipeline",)
+    CATEGORY = "mtb/audio"
+    FUNCTION = "load"
+
+    def load(self, model_size="tiny", download_missing=False):
+        """Load Whisper model and processor."""
+        whisper_dir = get_model_path("whisper")
+        tag = f"whisper-{model_size}"
+        model_dir = whisper_dir / tag
+
+        if not (whisper_dir.exists() or model_dir.exists()):
+            if not download_missing:
+                raise RuntimeError(
+                    "Models not found and download_missing=False"
+                )
+            else:
+                whisper_dir.mkdir(exist_ok=True)
+                model_dir.mkdir(exist_ok=True)
+
+                snapshot_download(
+                    repo_id=f"openai/{tag}",
+                    resume_download=True,
+                    ignore_patterns=["*.msgpack", "*.bin", "*.h5"],
+                    local_dir=model_dir.as_posix(),
+                    local_dir_use_symlinks=False,
+                )
+
+        device = get_torch_device()
+        log.debug(
+            f"Loading Whisper model {model_size} on {device} from {model_dir}"
+        )
+
+        processor = WhisperProcessor.from_pretrained(model_dir.as_posix())
+        model = WhisperForConditionalGeneration.from_pretrained(
+            model_dir.as_posix()
+        ).to(device)
+
+        model.eval()
+        model.requires_grad_(False)
+
+        return ({"processor": processor, "model": model},)
+
+
+class MTB_AudioToText(MtbAudio):
+    """Transcribe audio to text using Whisper."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "pipeline": ("WHISPER_PIPELINE",),
+                "audio": ("AUDIO",),
+                "language": (
+                    ["auto"]
+                    + sorted(
+                        [
+                            "en",
+                            "fr",
+                            "es",
+                            "de",
+                            "it",
+                            "pt",
+                            "nl",
+                            "ru",
+                            "zh",
+                            "ja",
+                            "ko",
+                        ]
+                    ),
+                    {"default": "auto"},
+                ),
+                "return_timestamps": ("BOOLEAN", {"default": True}),
+            },
+        }
+
+    RETURN_TYPES = ("STRING", "WHISPER_OUTPUT")
+    FUNCTION = "transcribe"
+    CATEGORY = "mtb/audio"
+
+    def transcribe(
+        self,
+        pipeline: WhisperPipeline,
+        audio: AudioTensor,
+        language="auto",
+        return_timestamps=True,
+    ):
+        """Transcribe audio to text using Whisper."""
+        processor = pipeline["processor"]
+        model = pipeline["model"]
+        device = model.device
+
+        audio = self.resample(audio, WHISPER_SAMPLE_RATE)
+
+        waveform = audio["waveform"]
+        log.debug(f"Processed waveform shape: {waveform.shape}")
+
+        # - Mono: [1, 1, samples] or [1, samples] or [samples]
+        # - Stereo: [1, 2, samples] or [2, samples] or [samples, 2]
+        if len(waveform.shape) == 3:
+            waveform = waveform.squeeze(0)
+
+        if len(waveform.shape) == 2:
+            if waveform.shape[0] == 2:  # [channels, samples]
+                waveform = waveform.mean(dim=0)
+            elif waveform.shape[1] == 2:  # [samples, channels]
+                waveform = waveform.mean(dim=1)
+            else:  # mono
+                waveform = waveform.squeeze(0)
+
+        sample_rate = audio["sample_rate"]
+        chunk_duration = 30
+        chunk_samples = chunk_duration * sample_rate
+        total_samples = waveform.shape[-1]
+        total_duration = total_samples / sample_rate
+
+        log.debug(f"Audio duration: {total_duration:.2f}s")
+
+        all_tokens = []
+        all_text = []
+        chunk_offsets = []
+
+        last_time = 0.0
+        accumulated_offset = 0.0
+
+        for chunk_start in range(0, total_samples, chunk_samples):
+            chunk_end = min(chunk_start + chunk_samples, total_samples)
+            chunk_waveform = waveform[chunk_start:chunk_end]
+            chunk_offset = chunk_start / sample_rate
+            chunk_offsets.append(chunk_offset)
+
+            log.debug(
+                f"Processing chunk {chunk_offset:.1f}s - {chunk_end / sample_rate:.1f}s"
+            )
+
+            max_length = model.config.max_length or 448
+            attention_mask = torch.ones((1, max_length))
+
+            input_features = processor(
+                chunk_waveform,
+                sampling_rate=sample_rate,
+                return_tensors="pt",
+            ).input_features.to(device)
+
+            with torch.no_grad():
+                predicted_ids = model.generate(
+                    input_features,
+                    attention_mask=attention_mask.to(device),
+                    task="transcribe",
+                    language=None if language == "auto" else language,
+                    return_timestamps=return_timestamps,
+                    no_repeat_ngram_size=3,
+                    num_beams=5,
+                    length_penalty=1.0,
+                    max_length=max_length,
+                )
+
+            chunk_tokens = processor.tokenizer.convert_ids_to_tokens(
+                predicted_ids[0]
+            )
+
+            adjusted_tokens = []
+            for token in chunk_tokens:
+                if token.startswith("<|") and token.endswith("|>"):
+                    try:
+                        time_str = token[2:-2]
+                        if time_str.replace(".", "").isdigit():
+                            time_val = float(time_str)
+
+                            # If this timestamp is less than the last one, we've started a new sequence
+                            if time_val < last_time:
+                                accumulated_offset += last_time
+
+                            adjusted_time = time_val + accumulated_offset
+                            adjusted_tokens.append(f"<|{adjusted_time:.2f}|>")
+                            last_time = time_val
+                        else:
+                            adjusted_tokens.append(token)
+                    except ValueError:
+                        adjusted_tokens.append(token)
+                else:
+                    adjusted_tokens.append(token)
+
+            all_tokens.extend(adjusted_tokens)
+            chunk_text = processor.batch_decode(
+                predicted_ids, skip_special_tokens=True
+            )[0]
+            all_text.append(chunk_text)
+
+        detected_language = "en"
+        if language == "auto":
+            try:
+                log.debug("Detecting language")
+                with torch.no_grad():
+                    first_chunk_features = processor(
+                        waveform[:chunk_samples],
+                        sampling_rate=sample_rate,
+                        return_tensors="pt",
+                    ).input_features.to(device)
+
+                    predicted_probs = model.detect_language(
+                        first_chunk_features
+                    )[0]
+                    language_token = processor.tokenizer.convert_ids_to_tokens(
+                        predicted_probs.argmax(-1).item()
+                    )
+                    detected_language = (
+                        language_token[2:-2]
+                        if language_token.startswith("<|")
+                        else "en"
+                    )
+                    log.debug(f"Detected language: {detected_language}")
+
+            except Exception as e:
+                log.warning(f"Language detection failed: {e}")
+
+        full_transcription = " ".join(all_text)
+
+        whisper_output = {
+            "text": full_transcription,
+            "language": detected_language,
+            "tokens": all_tokens,
+            "audio": audio,
+            "chunk_offsets": chunk_offsets,
+        }
+
+        return full_transcription, whisper_output
+
+
+class MTB_ProcessWhisperOutput:
+    """Process Whisper output into timestamped chunks."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "whisper_output": ("WHISPER_OUTPUT",),
+                "min_chunk_length": (
+                    "FLOAT",
+                    {"default": 0.0, "min": 0.0, "max": 10.0, "step": 0.1},
+                ),
+            },
+        }
+
+    RETURN_TYPES = ("STRING", "WHISPER_CHUNKS")
+    FUNCTION = "process"
+    CATEGORY = "mtb/audio"
+
+    def process(self, whisper_output, min_chunk_length=0.0):
+        """Process Whisper output into timestamped chunks."""
+        tokens = whisper_output["tokens"]
+        audio = whisper_output["audio"]
+        timestamp_tokens = []
+
+        audio_duration = audio["waveform"].shape[-1] / audio["sample_rate"]
+        log.debug(f"Audio duration: {audio_duration:.2f}s")
+
+        for i, token in enumerate(tokens):
+            if token.startswith("<|") and token.endswith("|>"):
+                try:
+                    time_str = token[2:-2]
+                    if time_str.replace(".", "").isdigit():
+                        time_val = float(time_str)
+                        if 0 <= time_val <= audio_duration:
+                            timestamp_tokens.append((i, time_val))
+                            log.debug(f"Token {i}: {time_val}")
+                except ValueError:
+                    continue
+
+        chunks = []
+        if len(timestamp_tokens) > 1:
+            for i in range(len(timestamp_tokens) - 1):
+                start_pos, start_time = timestamp_tokens[i]
+                end_pos, end_time = timestamp_tokens[i + 1]
+
+                if end_time - start_time < min_chunk_length:
+                    continue
+
+                chunk_tokens = tokens[start_pos + 1 : end_pos]
+                text = " ".join(
+                    t
+                    for t in chunk_tokens
+                    if not (t.startswith("<|") and t.endswith("|>"))
+                )
+
+                if text.strip():
+                    chunks.append(
+                        {
+                            "text": text.strip(),
+                            "timestamp": [start_time, end_time],
+                        }
+                    )
+
+        if timestamp_tokens:
+            start_pos, start_time = timestamp_tokens[-1]
+            if start_pos < len(tokens) - 1:
+                text = " ".join(
+                    t
+                    for t in tokens[start_pos + 1 :]
+                    if not (t.startswith("<|") and t.endswith("|>"))
+                )
+                if text.strip():
+                    if chunks:
+                        prev_chunk = chunks[-1]
+                        prev_duration = (
+                            prev_chunk["timestamp"][1]
+                            - prev_chunk["timestamp"][0]
+                        )
+                        end_time = min(
+                            start_time + prev_duration, audio_duration
+                        )
+                    else:
+                        end_time = audio_duration
+
+                    if (
+                        end_time > start_time
+                        and end_time - start_time >= min_chunk_length
+                    ):
+                        chunks.append(
+                            {
+                                "text": text.strip(),
+                                "timestamp": [start_time, end_time],
+                            }
+                        )
+
+        result = {
+            "text": whisper_output["text"],
+            "chunks": chunks,
+            "language": whisper_output["language"],
+        }
+
+        return whisper_output["text"], result
+
+
 class MTB_AudioCut(MtbAudio):
     """Basic audio cutter, values are in ms."""
 
@@ -98,7 +504,7 @@ class MTB_AudioCut(MtbAudio):
     CATEGORY = "mtb/audio"
     FUNCTION = "cut"
 
-    def cut(self, audio: AudioDict, length: float, offset: float):
+    def cut(self, audio: AudioTensor, length: float, offset: float):
         sample_rate = audio["sample_rate"]
         start_idx = int(offset * sample_rate / 1000)
         end_idx = min(
@@ -117,7 +523,6 @@ class MTB_AudioCut(MtbAudio):
 
 class MTB_AudioStack(MtbAudio):
     """Stack/Overlay audio inputs (dynamic inputs).
-
     - pad audios to the longest inputs.
     - resample audios to the highest sample rate in the inputs.
     - convert them all to stereo if one of the inputs is.
@@ -132,7 +537,7 @@ class MTB_AudioStack(MtbAudio):
     CATEGORY = "mtb/audio"
     FUNCTION = "stack"
 
-    def stack(self, **kwargs: AudioDict) -> tuple[AudioDict]:
+    def stack(self, **kwargs: AudioTensor) -> tuple[AudioTensor]:
         audios, is_stereo, max_rate = self.preprocess_audios(
             list(kwargs.values())
         )
@@ -163,7 +568,6 @@ class MTB_AudioStack(MtbAudio):
 
 class MTB_AudioSequence(MtbAudio):
     """Sequence audio inputs (dynamic inputs).
-
     - adding silence_duration between each segment
       can now also be negative to overlap the clips, safely bound
       to the the input length.
@@ -187,7 +591,7 @@ class MTB_AudioSequence(MtbAudio):
     CATEGORY = "mtb/audio"
     FUNCTION = "sequence"
 
-    def sequence(self, silence_duration: float, **kwargs: AudioDict):
+    def sequence(self, silence_duration: float, **kwargs: AudioTensor):
         audios, is_stereo, max_rate = self.preprocess_audios(
             list(kwargs.values())
         )
@@ -232,4 +636,287 @@ class MTB_AudioSequence(MtbAudio):
         )
 
 
-__nodes__ = [MTB_AudioSequence, MTB_AudioStack, MTB_AudioCut]
+class MTB_AudioResample(MtbAudio):
+    """Resample audio to a different sample rate."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "audio": ("AUDIO",),
+                "sample_rate": (
+                    "INT",
+                    {
+                        "default": 16000,
+                        "min": 1000,
+                        "max": 192000,
+                        "step": 100,
+                        "tooltip": "Target sample rate in Hz. Whisper requires 16000.",
+                    },
+                ),
+            }
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    RETURN_NAMES = ("resampled_audio",)
+    CATEGORY = "mtb/audio"
+    FUNCTION = "resample_audio"
+
+    def resample_audio(
+        self, audio: AudioTensor, sample_rate: int
+    ) -> tuple[AudioTensor]:
+        resampled = self.resample(audio, sample_rate)
+        return (resampled,)
+
+
+class MTB_AudioIsolateSpeaker(MtbAudio):
+    """Isolate or mute specific speakers using WhisperData"""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "audio": ("AUDIO",),
+                "whisper_data": ("WHISPER_CHUNKS",),
+                "target_speaker": ("STRING", {"default": "SPEAKER_00"}),
+                "mode": (["isolate", "mute"], {"default": "isolate"}),
+                "fade_ms": (
+                    "FLOAT",
+                    {
+                        "default": 100.0,
+                        "min": 0.0,
+                        "max": 1000.0,
+                        "step": 10,
+                        "tooltip": "Fade duration in milliseconds to avoid clicks",
+                    },
+                ),
+            },
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    RETURN_NAMES = ("processed_audio",)
+    CATEGORY = "mtb/audio"
+    FUNCTION = "process_audio"
+
+    def process_audio(
+        self,
+        audio: AudioTensor,
+        whisper_data: WhisperData,
+        target_speaker: str,
+        mode: str = "isolate",
+        fade_ms: float = 100.0,
+    ) -> tuple[AudioTensor]:
+        fade_samples = int((fade_ms / 1000.0) * audio["sample_rate"])
+
+        mask = (
+            torch.zeros_like(audio["waveform"])
+            if mode == "isolate"
+            else torch.ones_like(audio["waveform"])
+        )
+
+        for chunk in whisper_data["chunks"]:
+            if not chunk.get("speaker"):
+                continue
+
+            speaker_present = target_speaker in chunk["speaker"]
+            if (mode == "isolate" and speaker_present) or (
+                mode == "mute" and not speaker_present
+            ):
+                start_sample = int(
+                    chunk["timestamp"][0] * audio["sample_rate"]
+                )
+                end_sample = int(chunk["timestamp"][1] * audio["sample_rate"])
+
+                mask[:, start_sample:end_sample] = 1.0
+
+        if fade_samples > 0:
+            fade = torch.linspace(0, 1, fade_samples)
+
+            transitions = torch.where(mask[0, 1:] != mask[0, :-1])[0] + 1
+
+            for trans_idx in transitions:
+                if (
+                    trans_idx >= fade_samples
+                    and trans_idx <= mask.shape[1] - fade_samples
+                ):
+                    if mask[0, trans_idx] == 1:
+                        mask[:, trans_idx : trans_idx + fade_samples] *= fade
+                    else:
+                        mask[:, trans_idx - fade_samples : trans_idx] *= (
+                            fade.flip(0)
+                        )
+
+        processed_waveform = audio["waveform"] * mask
+
+        return (
+            {
+                "sample_rate": audio["sample_rate"],
+                "waveform": processed_waveform,
+            },
+        )
+
+
+class MTB_ProcessWhisperDiarization:
+    """Process Whisper chunks with speaker diarization using either pyannote or NeMo."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "whisper_chunks": ("WHISPER_CHUNKS",),
+                "audio": ("AUDIO",),
+                "backend": (["pyannote", "nemo"], {"default": "pyannote"}),
+                "num_speakers": (
+                    "INT",
+                    {"default": 2, "min": 1, "max": 10, "step": 1},
+                ),
+            },
+            "optional": {
+                "device": (["cuda", "cpu"], {"default": "cuda"}),
+            },
+        }
+
+    RETURN_TYPES = ("WHISPER_CHUNKS",)
+    FUNCTION = "process"
+    CATEGORY = "mtb/audio"
+
+    def process_pyannote(self, audio, num_speakers, device):
+        """Process audio using pyannote backend."""
+        try:
+            from pyannote.audio import Pipeline
+            from pyannote.audio.pipelines.utils.hook import ProgressHook
+        except ImportError:
+            raise ImportError(
+                "pyannote.audio not found. Install with: pip install pyannote.audio"
+            )
+
+        pipeline = Pipeline.from_pretrained(
+            "pyannote/speaker-diarization-3.1", use_auth_token=None
+        )
+        pipeline.to(torch.device(device))
+        with ProgressHook() as hook:
+            diarization = pipeline(
+                {
+                    "waveform": audio["waveform"][0],
+                    "sample_rate": audio["sample_rate"],
+                },
+                num_speakers=num_speakers,
+                hook=hook,
+            )
+
+        speaker_segments = []
+        for turn, _, speaker in diarization.itertracks(yield_label=True):
+            speaker_segments.append(
+                {
+                    "start": turn.start,
+                    "end": turn.end,
+                    "speaker": speaker,
+                }
+            )
+
+        return speaker_segments
+
+    def process_nemo(self, audio, num_speakers, device):
+        """Process audio using NeMo backend."""
+        try:
+            import nemo.collections.asr as nemo_asr
+        except ImportError:
+            raise ImportError(
+                "NeMo not found. Install with: pip install nemo_toolkit[asr]"
+            )
+
+        model = nemo_asr.models.ClusteringDiarizer.from_pretrained(
+            "nvidia/speakerverification_en_titanet_large"
+        ).to(device)
+
+        diarization = model.diarize(
+            audio=audio["waveform"][0],
+            sample_rate=audio["sample_rate"],
+            num_speakers=num_speakers,
+        )
+
+        speaker_segments = []
+        for segment in diarization:
+            speaker_segments.append(
+                {
+                    "start": segment["start"],
+                    "end": segment["end"],
+                    "speaker": f"SPEAKER_{segment['speaker']}",
+                }
+            )
+
+        return speaker_segments
+
+    def process(
+        self,
+        whisper_chunks,
+        audio,
+        backend="pyannote",
+        num_speakers=2,
+        device="cuda",
+    ):
+        if backend == "pyannote":
+            speaker_segments = self.process_pyannote(
+                audio, num_speakers, device
+            )
+        else:  # nemo
+            speaker_segments = self.process_nemo(audio, num_speakers, device)
+
+        for chunk in whisper_chunks["chunks"]:
+            chunk_start, chunk_end = chunk["timestamp"]
+            chunk_speakers = set()
+            for segment in speaker_segments:
+                if (
+                    segment["start"] <= chunk_end
+                    and segment["end"] >= chunk_start
+                ):
+                    chunk_speakers.add(segment["speaker"])
+
+            if chunk_speakers:
+                chunk["speaker"] = list(chunk_speakers)[0]
+            else:
+                chunk["speaker"] = "unknown"
+
+        return (whisper_chunks,)
+
+
+class MTB_AudioDuration:
+    """Get audio duration in milliseconds."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "audio": ("AUDIO",),
+            },
+        }
+
+    RETURN_TYPES = ("INT",)
+    RETURN_NAMES = ("duration_ms",)
+    FUNCTION = "get_duration"
+    CATEGORY = "mtb/audio"
+
+    def get_duration(self, audio):
+        waveform = audio["waveform"]
+        sample_rate = audio["sample_rate"]
+
+        duration_ms = int((waveform.shape[-1] / sample_rate) * 1000)
+        log.debug(
+            f"Audio duration: {duration_ms}ms ({duration_ms / 1000:.2f}s)"
+        )
+
+        return (duration_ms,)
+
+
+__nodes__ = [
+    MTB_AudioSequence,
+    MTB_AudioStack,
+    MTB_AudioCut,
+    MTB_AudioResample,
+    MTB_AudioIsolateSpeaker,
+    MTB_LoadWhisper,
+    MTB_AudioToText,
+    MTB_ProcessWhisperOutput,
+    MTB_ProcessWhisperDiarization,
+    MTB_AudioDuration,
+]

nodes/batch.py

@@ -1,12 +1,18 @@
+import os
+import random
 from io import BytesIO
+from pathlib import Path
+from typing import Literal
 
+import comfy.utils
 import cv2
+import folder_paths
 import numpy as np
 import torch
 from PIL import Image
 
 from ..log import log
-from ..utils import EASINGS, apply_easing, pil2tensor
+from ..utils import EASINGS, apply_easing, glob_multiple, pil2tensor
 from .transform import MTB_TransformImage
 
 
@@ -46,7 +52,7 @@ class MTB_BatchFloatMath:
         for v in vals:
             if len(v) != ref_count:
                 raise ValueError(
-                    f"All values must have the same length (current: {len(v)}, ref: {ref_count}"
+                    f"All values must have the same length (current: {len(v)}, ref: {ref_count})"
                 )
 
         match operation:
@@ -169,6 +175,124 @@ class MTB_BatchTimeWrap:
         return (warped_tensor, interpolated_curve)
 
 
+class MTB_ImageBatchToSublist:
+    """
+    # Image Batch To Sublist 🔄
+
+    Splits a large batched tensor into smaller sub-batches for memory-efficient processing.
+    """
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "sub_batch_size": (
+                    "INT",
+                    {"default": 1, "min": 1, "max": 1000, "step": 1},
+                ),
+            },
+            "optional": {
+                "image": ("IMAGE",),
+                "mask": ("MASK",),
+            },
+        }
+
+    RETURN_TYPES = ("IMAGE", "MASK", "INT")
+    RETURN_NAMES = ("image_list", "mask_list", "item_count")
+
+    OUTPUT_IS_LIST = (True, True)
+    FUNCTION = "split_batch"
+    CATEGORY = "batch_processing"
+
+    def split_batch(
+        self,
+        sub_batch_size: int,
+        image: torch.Tensor | None = None,
+        mask: torch.Tensor | None = None,
+    ):
+        if image is None and mask is None:
+            raise ValueError(
+                "You must either pass mask or image, none received"
+            )
+
+        image_count = 0
+        if image is not None:
+            image_count = image.size(0)
+
+        mask_count = 0
+        if mask is not None:
+            mask_count = mask.size(0)
+
+        if image_count > 0 and mask_count > 0 and mask_count != image_count:
+            raise ValueError(
+                f"When providing image and mask, batch size must match (got {mask.size(0)} mask and {image.size(0)} images)"
+            )
+
+        batch_size = max(image_count, mask_count)
+
+        num_full_batches = batch_size // sub_batch_size
+        im_batches = []
+        mask_batches = []
+
+        for i in range(num_full_batches):
+            start_idx = i * sub_batch_size
+            end_idx = start_idx + sub_batch_size
+            if image_count > 0:
+                im_batches.append(image[start_idx:end_idx, ...])
+
+            if mask_count > 0:
+                mask_batches.append(mask[start_idx:end_idx, ...])
+
+        if batch_size % sub_batch_size != 0:
+            remaining_start = num_full_batches * sub_batch_size
+            if image_count > 0:
+                im_batches.append(image[remaining_start:, ...])
+
+            if mask_count > 0:
+                mask_batches.append(mask[remaining_start:, ...])
+
+        return (im_batches, mask_batches, len(im_batches))
+
+
+class MTB_SublistToImageBatch:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "tensors": ("IMAGE",),
+            }
+        }
+
+    INPUT_IS_LIST = True
+    RETURN_TYPES = ("IMAGE",)
+    FUNCTION = "merge_batches"
+    CATEGORY = "batch_processing"
+    DOCUMENTATION = """# Sublist to Image Batch 🔄
+
+Merges a list of sub-batched tensors back into a single large batch.
+"""
+
+    def merge_batches(self, tensors: list[torch.Tensor]):
+        if len(tensors) <= 1:
+            return (tensors[0],)
+
+        result = tensors[0]
+
+        for next_tensor in tensors[1:]:
+            if result.shape[1:] != next_tensor.shape[1:]:
+                next_tensor = comfy.utils.common_upscale(
+                    next_tensor.movedim(-1, 1),
+                    result.shape[2],
+                    result.shape[1],
+                    "lanczos",
+                    "center",
+                ).movedim(1, -1)
+
+            result = torch.cat((result, next_tensor), dim=0)
+
+        return (result,)
+
+
 class MTB_BatchMake:
     """Simply duplicates the input frame as a batch"""
 
@@ -178,18 +302,22 @@ class MTB_BatchMake:
             "required": {
                 "image": ("IMAGE",),
                 "count": ("INT", {"default": 1}),
-            }
+            },
+            "optional": {"mask": ("MASK",)},
         }
 
     RETURN_TYPES = ("IMAGE",)
     FUNCTION = "generate_batch"
     CATEGORY = "mtb/batch"
 
-    def generate_batch(self, image: torch.Tensor, count):
+    def generate_batch(self, image: torch.Tensor, count, mask=None):
         if len(image.shape) == 3:
             image = image.unsqueeze(0)
 
-        return (image.repeat(count, 1, 1, 1),)
+        return (
+            image.repeat(count, 1, 1, 1),
+            mask.repeat(count, 1, 1) if mask else mask,
+        )
 
 
 class MTB_BatchShape:
@@ -374,8 +502,14 @@ class MTB_BatchFloat:
                     {"default": "Steps"},
                 ),
                 "count": ("INT", {"default": 2}),
-                "min": ("FLOAT", {"default": 0.0, "step": 0.001}),
-                "max": ("FLOAT", {"default": 1.0, "step": 0.001}),
+                "min": (
+                    "FLOAT",
+                    {"default": 0.0, "min": -1e4, "max": 1e4, "step": 0.001},
+                ),
+                "max": (
+                    "FLOAT",
+                    {"default": 1.0, "min": -1e4, "max": 1e4, "step": 0.001},
+                ),
                 "easing": (
                     [
                         "Linear",
@@ -410,7 +544,14 @@ class MTB_BatchFloat:
     RETURN_TYPES = ("FLOATS",)
     CATEGORY = "mtb/batch"
 
-    def set_floats(self, mode, count, min, max, easing):
+    def set_floats(
+        self,
+        mode: Literal["Steps"] | Literal["Single"] = "Steps",
+        count: int = 1,
+        min: float = 0.0,  # noqa: A002
+        max: float = 1.0,  # noqa: A002
+        easing: str = "Linear",
+    ):
         if mode == "Steps" and count == 1:
             raise ValueError(
                 "Steps mode requires at least a count of 2 values"
@@ -429,6 +570,210 @@ class MTB_BatchFloat:
         return (keyframes,)
 
 
+class MTB_BatchSequencePlus:
+    """Sequences multiple image batches with transition effects."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "transition": (
+                    [
+                        "none",
+                        "crossfade",
+                        "slide_left",
+                        "slide_right",
+                        "slide_up",
+                        "slide_down",
+                        "wipe_left",
+                        "wipe_right",
+                        "wipe_up",
+                        "wipe_down",
+                        "band_wipe_h",
+                        "band_wipe_v",
+                    ],
+                    {"default": "none"},
+                ),
+                "overlap_frames": (
+                    "INT",
+                    {"default": 0, "min": 0, "max": 120, "step": 1},
+                ),
+                "reverse": ("BOOLEAN", {"default": False}),
+            }
+        }
+
+    RETURN_TYPES = ("IMAGE",)
+    FUNCTION = "sequence_batches"
+    CATEGORY = "mtb/batch"
+
+    def apply_transition(
+        self,
+        frame1: torch.Tensor,
+        frame2: torch.Tensor,
+        transition: str,
+        progress: float,
+    ):
+        """Apply transition effect between two frames."""
+        if transition == "none":
+            return frame1 if progress < 0.5 else frame2
+
+        elif transition == "crossfade":
+            return frame1 * (1 - progress) + frame2 * progress
+
+        elif transition.startswith("slide_"):
+            h, w = frame1.shape[1:3]
+            if transition == "slide_left":
+                offset = int(w * progress)
+                frame2 = torch.roll(frame2, shifts=-offset, dims=2)
+            elif transition == "slide_right":
+                offset = int(w * progress)
+                frame2 = torch.roll(frame2, shifts=offset, dims=2)
+            elif transition == "slide_up":
+                offset = int(h * progress)
+                frame2 = torch.roll(frame2, shifts=-offset, dims=1)
+            elif transition == "slide_down":
+                offset = int(h * progress)
+                frame2 = torch.roll(frame2, shifts=offset, dims=1)
+            return frame1 * (1 - progress) + frame2 * progress
+
+        elif transition.startswith("wipe_"):
+            h, w = frame1.shape[1:3]
+            mask = torch.zeros_like(frame1)
+            if transition == "wipe_left":
+                edge = int(w * progress)
+                mask[:, :, :edge, :] = 1
+            elif transition == "wipe_right":
+                edge = int(w * (1 - progress))
+                mask[:, :, edge:, :] = 1
+            elif transition == "wipe_up":
+                edge = int(h * progress)
+                mask[:, :edge, :, :] = 1
+            elif transition == "wipe_down":
+                edge = int(h * (1 - progress))
+                mask[:, edge:, :, :] = 1
+            return frame1 * (1 - mask) + frame2 * mask
+
+        elif transition.startswith("band_wipe_"):
+            h, w = frame1.shape[1:3]
+            mask = torch.zeros_like(frame1)
+            num_bands = 10  # Number of bands
+
+            if transition == "band_wipe_h":
+                band_width = w / num_bands
+                for i in range(num_bands):
+                    edge = int((w * progress) - (i * band_width))
+                    start = int(i * band_width)
+                    end = int(min(start + edge, (i + 1) * band_width))
+                    if end > start:
+                        mask[:, :, start:end, :] = 1
+            else:  # band_wipe_v
+                band_height = h / num_bands
+                for i in range(num_bands):
+                    edge = int((h * progress) - (i * band_height))
+                    start = int(i * band_height)
+                    end = int(min(start + edge, (i + 1) * band_height))
+                    if end > start:
+                        mask[:, start:end, :, :] = 1
+
+            return frame1 * (1 - mask) + frame2 * mask
+
+        return frame1
+
+    def sequence_batches(
+        self, transition: str, overlap_frames: int, reverse: bool, **kwargs
+    ):
+        images: list[torch.Tensor] = list(kwargs.values())
+
+        if reverse:
+            images = images[::-1]
+
+        processed_images: list[torch.Tensor] = []
+        for img in images:
+            if len(img.shape) == 3:
+                img = img.unsqueeze(0)
+            processed_images.append(img)
+
+        if overlap_frames == 0 or transition == "none":
+            return (torch.cat(processed_images, dim=0),)
+
+        result_frames: list[torch.Tensor] = []
+
+        if len(processed_images) > 0:
+            result_frames.extend(
+                list(processed_images[0][: -overlap_frames // 2])
+            )
+
+        for i in range(1, len(processed_images)):
+            prev_batch = processed_images[i - 1]
+            curr_batch = processed_images[i]
+
+            prev_frames = min(overlap_frames // 2, len(prev_batch))
+            next_frames = min(overlap_frames // 2, len(curr_batch))
+            total_overlap = prev_frames + next_frames
+
+            if total_overlap < 2:
+                # when not enough frames for transition, just concatenate
+                result_frames.extend(list(prev_batch[-prev_frames:]))
+                result_frames.extend(list(curr_batch[:next_frames]))
+                continue
+
+            for t in range(total_overlap):
+                progress = t / (total_overlap - 1)
+
+                prev_idx = (
+                    len(prev_batch) - prev_frames + min(t, prev_frames - 1)
+                )
+                next_idx = max(0, t - prev_frames)
+
+                transition_frame = self.apply_transition(
+                    prev_batch[prev_idx : prev_idx + 1],
+                    curr_batch[next_idx : next_idx + 1],
+                    transition,
+                    progress,
+                )
+                result_frames.append(transition_frame[0])
+
+            if i < len(processed_images) - 1:
+                result_frames.extend(
+                    list(curr_batch[next_frames : -overlap_frames // 2])
+                )
+            else:
+                result_frames.extend(list(curr_batch[next_frames:]))
+
+        result = torch.stack(result_frames, dim=0)
+
+        return (result,)
+
+
+class MTB_BatchSequence:
+    """Sequences multiple image batches one after another"""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "reverse": ("BOOLEAN", {"default": False}),
+            }
+        }
+
+    RETURN_TYPES = ("IMAGE",)
+    FUNCTION = "sequence_batches"
+    CATEGORY = "mtb/batch"
+
+    def sequence_batches(self, reverse: bool, **kwargs):
+        images = list(kwargs.values())
+        if reverse:
+            images = images[::-1]
+
+        processed = []
+        for img in images:
+            if len(img.shape) == 3:
+                img = img.unsqueeze(0)
+            processed.append(img)
+
+        return (torch.cat(processed, dim=0),)
+
+
 class MTB_BatchMerge:
     """Merges multiple image batches with different frame counts"""
 
@@ -505,6 +850,13 @@ class MTB_Batch2dTransform:
                 "zoom": ("FLOATS",),
                 "angle": ("FLOATS",),
                 "shear": ("FLOATS",),
+                "use_normalized": (
+                    "BOOLEAN",
+                    {
+                        "default": False,
+                        "tooltip": "If true, transform values will be scaled to image dimensions.",
+                    },
+                ),
             },
         }
 
@@ -533,6 +885,7 @@ class MTB_Batch2dTransform:
         zoom: list[float] | None = None,
         angle: list[float] | None = None,
         shear: list[float] | None = None,
+        use_normalized: bool = False,
     ):
         if all(
             self.get_num_elements(param) <= 0
@@ -584,6 +937,7 @@ class MTB_Batch2dTransform:
                 keyframes["shear"][i],
                 border_handling,
                 constant_color,
+                use_normalized=use_normalized,
             )[0]
             for i in range(image.shape[0])
         ]
@@ -711,7 +1065,9 @@ class MTB_PlotBatchFloat:
         ax.set_xlim(1, max_length)  # Set X-axis limits
         np.random.seed(seed)
         colors = np.random.rand(len(kwargs), 3)  # Generate random RGB values
-        for color, (label, values) in zip(colors, kwargs.items()):
+        for color, (label, values) in zip(
+            colors, kwargs.items(), strict=False
+        ):
             ax.plot(x_values[: len(values)], values, label=label, color=color)
         ax.legend(
             title="Legend",
@@ -1025,18 +1381,163 @@ class MTB_BatchShake:
         return (shaken_images, x_translations, y_translations, rotations)
 
 
+class MTB_BatchFromFolder:
+    """Load images from a folder with options for latest, oldest, or random selection."""
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "enable": (
+                    "BOOLEAN",
+                    {
+                        "default": True,
+                        "tooltip": "Enable or disable the node. If disabled, returns passthrough_image or an empty tensor.",
+                    },
+                ),
+                "folder_path": (
+                    "STRING",
+                    {
+                        "default": "",
+                        "tooltip": "Path to the folder containing images. Relative paths are resolved to the ComfyUI output directory.",
+                    },
+                ),
+                "mode": (
+                    ["latest", "oldest", "random"],
+                    {
+                        "default": "latest",
+                        "tooltip": "How to select images: latest, oldest, or random.",
+                    },
+                ),
+                "count": (
+                    "INT",
+                    {
+                        "default": 10,
+                        "min": 1,
+                        "max": 1000,
+                        "tooltip": "Number of images to load from the folder.",
+                    },
+                ),
+                "filter": (
+                    "STRING",
+                    {
+                        "default": "*",
+                        "tooltip": "Glob filter for image filenames (e.g. *.png).",
+                    },
+                ),
+            },
+            "optional": {
+                "passthrough_image": (
+                    "IMAGE",
+                    {
+                        "tooltip": "If provided and node is disabled, this image is passed through instead of returning an empty tensor."
+                    },
+                ),
+            },
+        }
+
+    RETURN_TYPES = ("IMAGE",)
+    RETURN_NAMES = ("images",)
+    CATEGORY = "mtb/batch"
+    FUNCTION = "load_from_folder"
+
+    def load_from_folder(
+        self,
+        enable: bool,
+        folder_path: str,
+        mode: str,
+        count: int,
+        filter: str,
+        passthrough_image=None,
+    ):
+        """Load images from a folder with the specified selection mode."""
+        if not enable:
+            if passthrough_image is not None:
+                log.debug(
+                    "MTB_BatchFromFolder: Using passthrough image (disabled)"
+                )
+                return (passthrough_image,)
+            log.debug(
+                "MTB_BatchFromFolder: Disabled and no passthrough_image provided, returning empty tensor"
+            )
+            return (torch.zeros(0, 0, 0, 3),)
+
+        path_obj = Path(folder_path)
+        if not path_obj.is_absolute():
+            output_dir = Path(folder_paths.get_output_directory())
+            path_obj = output_dir / folder_path
+            path_obj = path_obj.resolve()
+
+        if not path_obj.exists():
+            log.error(f"Folder path does not exist: {path_obj}")
+            return (torch.zeros(0, 0, 0, 3),)
+
+        if not path_obj.is_dir():
+            log.error(f"Path is not a directory: {path_obj}")
+            return (torch.zeros(0, 0, 0, 3),)
+
+        patterns = [filter] if filter else ["*"]
+        files = glob_multiple(path_obj, patterns)
+
+        image_extensions = [".png", ".jpg", ".jpeg", ".bmp", ".webp", ".tiff"]
+        image_files = [
+            f for f in files if f.suffix.lower() in image_extensions
+        ]
+
+        if not image_files:
+            log.warning(
+                f"No image files found in {path_obj} with filter {filter}"
+            )
+            return (torch.zeros(0, 0, 0, 3),)
+
+        if mode == "latest":
+            image_files.sort(key=lambda x: os.path.getmtime(x), reverse=True)
+        elif mode == "oldest":
+            image_files.sort(key=lambda x: os.path.getmtime(x))
+        elif mode == "random":
+            random.shuffle(image_files)
+
+        selected_files = image_files[:count]
+
+        if len(selected_files) < count:
+            log.warning(
+                f"Requested {count} images but only found {len(selected_files)}"
+            )
+
+        loaded_images = []
+        for file_path in selected_files:
+            try:
+                img = Image.open(file_path)
+                if img.mode != "RGB":
+                    img = img.convert("RGB")
+                loaded_images.append(img)
+            except Exception as e:
+                log.error(f"Error loading image {file_path}: {e}")
+
+        if not loaded_images:
+            log.error("Failed to load any images")
+            return (torch.zeros(0, 0, 0, 3),)
+
+        return (pil2tensor(loaded_images),)
+
+
 __nodes__ = [
-    MTB_BatchFloat,
     MTB_Batch2dTransform,
-    MTB_BatchShape,
-    MTB_BatchMake,
+    MTB_BatchFloat,
     MTB_BatchFloatAssemble,
     MTB_BatchFloatFill,
-    MTB_BatchFloatNormalize,
-    MTB_BatchMerge,
-    MTB_BatchShake,
-    MTB_PlotBatchFloat,
-    MTB_BatchTimeWrap,
     MTB_BatchFloatFit,
     MTB_BatchFloatMath,
+    MTB_BatchFloatNormalize,
+    MTB_BatchFromFolder,
+    MTB_BatchMake,
+    MTB_BatchMerge,
+    MTB_BatchSequence,
+    MTB_BatchSequencePlus,
+    MTB_BatchShake,
+    MTB_BatchShape,
+    MTB_BatchTimeWrap,
+    MTB_PlotBatchFloat,
+    MTB_SublistToImageBatch,
+    MTB_ImageBatchToSublist,
 ]

nodes/crop.py

@@ -78,14 +78,17 @@ class MTB_UpscaleBboxBy:
         self, bbox: tuple[int, int, int, int], scale: float
     ) -> tuple[tuple[int, int, int, int]]:
         x, y, width, height = bbox
-        # scaled = (x * scale, y * scale, width * scale, height * scale)
-        scaled = (
-            int(x * scale),
-            int(y * scale),
-            int(width * scale),
-            int(height * scale),
-        )
 
+        center_x = x + width // 2
+        center_y = y + height // 2
+
+        new_width = int(width * scale)
+        new_height = int(height * scale)
+
+        new_x = center_x - new_width // 2
+        new_y = center_y - new_height // 2
+
+        scaled = (new_x, new_y, new_width, new_height)
         return (scaled,)
 
 
@@ -372,6 +375,62 @@ class MTB_Uncrop:
         return (pil2tensor(out_images),)
 
 
+class MTB_BBoxForceDimensions:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "bbox": ("BBOX",),
+                "width": ("INT", {"default": 512, "min": 1, "max": 8192}),
+                "height": ("INT", {"default": 512, "min": 1, "max": 8192}),
+            },
+            "optional": {
+                "image": ("IMAGE",),
+            },
+        }
+
+    CATEGORY = "mtb/crop"
+    RETURN_TYPES = ("BBOX",)
+    FUNCTION = "force_dimensions"
+
+    def force_dimensions(
+        self,
+        bbox: tuple[int, int, int, int],
+        width: int,
+        height: int,
+        image: torch.Tensor = None,
+    ) -> tuple[tuple[int, int, int, int]]:
+        x, y, curr_width, curr_height = bbox
+
+        center_x = x + curr_width // 2
+        center_y = y + curr_height // 2
+
+        new_x = center_x - width // 2
+        new_y = center_y - height // 2
+
+        if image is not None:
+            img_height, img_width = image.shape[1:3]
+            x_overflow = max(0, new_x + width - img_width) + min(0, new_x)
+            y_overflow = max(0, new_y + height - img_height) + min(0, new_y)
+            if width > img_width or height > img_height:
+                x_exceed = width - img_width if width > img_width else 0
+                y_exceed = height - img_height if height > img_height else 0
+                raise ValueError(
+                    f"Target bbox dimensions ({width}x{height}) exceed image bounds ({img_width}x{img_height}) "
+                    f"by {x_exceed}px horizontally and {y_exceed}px vertically"
+                )
+
+            if x_overflow > 0 or x_overflow < 0:
+                new_x -= x_overflow
+
+            if y_overflow > 0:
+                new_y -= y_overflow
+            elif y_overflow < 0:
+                new_y -= y_overflow  # Add the negative overflow
+
+        return ((int(new_x), int(new_y), width, height),)
+
+
 __nodes__ = [
     MTB_BboxFromMask,
     MTB_Bbox,
@@ -379,4 +438,5 @@ __nodes__ = [
     MTB_Uncrop,
     MTB_SplitBbox,
     MTB_UpscaleBboxBy,
+    MTB_BBoxForceDimensions,
 ]

nodes/debug.py

@@ -2,7 +2,6 @@ import base64
 import io
 import json
 from pathlib import Path
-from typing import Optional
 
 import folder_paths
 import torch
@@ -11,13 +10,66 @@ from ..log import log
 from ..utils import tensor2pil
 
 
+def get_detailed_type_info(obj):
+    type_info = []
+
+    type_name = type(obj).__name__
+    type_info.append(f"Type: {type_name}")
+
+    if isinstance(obj, torch.Tensor):
+        type_info.extend(
+            [
+                f"Shape: {obj.shape}",
+                f"Dtype: {obj.dtype}",
+                f"Device: {obj.device}",
+                f"Requires grad: {obj.requires_grad}",
+                f"Stride: {obj.stride()}",
+                f"Contiguous: {obj.is_contiguous()}",
+            ]
+        )
+    elif isinstance(obj, (list, tuple)):
+        type_info.extend(
+            [
+                f"Length: {len(obj)}",
+                f"Container type: {type_name}",
+            ]
+        )
+        if obj:
+            type_info.append(f"Element type: {type(obj[0]).__name__}")
+    elif isinstance(obj, dict):
+        type_info.extend(
+            [
+                f"Length: {len(obj)}",
+                f"Keys: {list(obj.keys())}",
+            ]
+        )
+    elif hasattr(obj, "__dict__"):
+        attributes = [attr for attr in dir(obj) if not attr.startswith("_")]
+        type_info.append(f"Attributes: {attributes}")
+
+    return type_info
+
+
 # region processors
-def process_tensor(tensor):
+def process_tensor(tensor: torch.Tensor, as_type=False):
     log.debug(f"Tensor: {tensor.shape}")
 
+    if as_type:
+        return {
+            "text": [f"Tensor of shape {tensor.shape} of type {tensor.dtype}"]
+        }
+
+    is_mask = len(tensor.shape) == 3
+
+    if is_mask:
+        tensor = tensor.unsqueeze(-1).repeat(1, 1, 1, 3)
+
     image = tensor2pil(tensor)
     b64_imgs = []
     for im in image:
+        if is_mask:
+            im = im.convert("L")
+
         buffered = io.BytesIO()
         im.save(buffered, format="PNG")
         b64_imgs.append(
@@ -28,11 +80,16 @@ def process_tensor(tensor):
     return {"b64_images": b64_imgs}
 
 
-def process_list(anything):
+def process_list(anything, as_type=False):
     text = []
     if not anything:
         return {"text": []}
 
+    if as_type:
+        type_info = get_detailed_type_info(anything)
+        type_info.extend(get_detailed_type_info(anything[0]))
+        return {"text": type_info}
+
     first_element = anything[0]
     if (
         isinstance(first_element, list)
@@ -54,25 +111,41 @@ def process_list(anything):
     return {"text": text}
 
 
-def process_dict(anything):
+def process_dict(anything, as_type=False):
     text = []
+    if as_type:
+        return {"text": get_detailed_type_info(anything)}
+
     if "samples" in anything:
         is_empty = (
             "(empty)" if torch.count_nonzero(anything["samples"]) == 0 else ""
         )
         text.append(f"Latent Samples: {anything['samples'].shape} {is_empty}")
 
+    elif "waveform" in anything:
+        is_empty = (
+            "(empty) " if torch.count_nonzero(anything["samples"]) == 0 else ""
+        )
+
+        text.append(
+            f"Audio Samples: {anything['waveform'].shape}{is_empty} | sample rate {anything['sample_rate']}"
+        )
+
     else:
+        log.debug(f"Unhandled dict: {anything.keys()}")
         text.append(json.dumps(anything, indent=2))
 
     return {"text": text}
 
 
-def process_bool(anything):
+def process_bool(anything, as_type=False):
     return {"text": ["True" if anything else "False"]}
 
 
-def process_text(anything):
+def process_text(anything, as_type=False):
+    if as_type:
+        return {"text": get_detailed_type_info(anything)}
+
     return {"text": [str(anything)]}
 
 
@@ -89,6 +162,7 @@ class MTB_Debug:
     def INPUT_TYPES(cls):
         return {
             "required": {"output_to_console": ("BOOLEAN", {"default": False})},
+            "optional": {"as_detailed_types": ("BOOLEAN", {"default": False})},
         }
 
     RETURN_TYPES = ()
@@ -96,29 +170,25 @@ class MTB_Debug:
     CATEGORY = "mtb/debug"
     OUTPUT_NODE = True
 
-    def do_debug(self, output_to_console: bool, **kwargs):
-        output = {
-            "ui": {"b64_images": [], "text": []},
-            # "result": ("A"),
-        }
+    def do_debug(
+        self, output_to_console: bool, as_detailed_types: bool, **kwargs
+    ):
+        output = {"ui": {"items": []}}
 
-        processors = {
-            torch.Tensor: process_tensor,
-            list: process_list,
-            dict: process_dict,
-            bool: process_bool,
-        }
         if output_to_console:
             for k, v in kwargs.items():
                 log.info(f"{k}: {v}")
 
-        for anything in kwargs.values():
+        for input_name, anything in kwargs.items():
             processor = processors.get(type(anything), process_text)
 
-            processed_data = processor(anything)
+            processed = processor(anything, as_detailed_types)
 
-            for ui_key, ui_value in processed_data.items():
-                output["ui"][ui_key].extend(ui_value)
+            item = {
+                "input": input_name,
+                **processed,
+            }
+            output["ui"]["items"].append(item)
 
         return output
 
@@ -154,9 +224,9 @@ class MTB_SaveTensors:
     def save(
         self,
         filename_prefix,
-        image: Optional[torch.Tensor] = None,
-        mask: Optional[torch.Tensor] = None,
-        latent: Optional[torch.Tensor] = None,
+        image: torch.Tensor | None = None,
+        mask: torch.Tensor | None = None,
+        latent: torch.Tensor | None = None,
     ):
         (
             full_output_folder,
@@ -188,4 +258,11 @@ class MTB_SaveTensors:
         return f"{filename_prefix}_{counter:05}"
 
 
+processors = {
+    torch.Tensor: process_tensor,
+    list: process_list,
+    dict: process_dict,
+    bool: process_bool,
+}
+
 __nodes__ = [MTB_Debug, MTB_SaveTensors]

nodes/deep_bump.py

@@ -2,13 +2,16 @@ import tempfile
 from pathlib import Path
 
 import numpy as np
+
+# torch must be imported prior to onnx for the CUDAProvider.
+import torch  # isort:skip
 import onnxruntime as ort
-import torch
 from PIL import Image
 
 from ..errors import ModelNotFound
 from ..log import mklog
 from ..utils import (
+    download_model,
     get_model_path,
     tensor2pil,
     tiles_infer,
@@ -23,7 +26,12 @@ log = mklog(__name__)
 
 # - COLOR to NORMALS
 def color_to_normals(
-    color_img, overlap, progress_callback, *, save_temp=False
+    color_img,
+    overlap,
+    progress_callback,
+    *,
+    save_temp=False,
+    auto_download=False,
 ):
     """Compute a normal map from the given color map.
 
@@ -67,7 +75,13 @@ def color_to_normals(
     log.debug("DeepBump Color → Normals : loading model")
     model = get_model_path("deepbump", "deepbump256.onnx")
     if not model or not model.exists():
-        raise ModelNotFound(f"deepbump ({model})")
+        if not auto_download:
+            raise ModelNotFound(f"deepbump ({model})")
+        log.debug("Downloading models...")
+        download_model(
+            "https://github.com/HugoTini/DeepBump/raw/master/deepbump256.onnx",
+            "deepbump",
+        )
 
     providers = [
         "TensorrtExecutionProvider",
@@ -351,6 +365,9 @@ class MTB_DeepBump:
                 ),
                 "normals_to_height_seamless": ("BOOLEAN", {"default": True}),
             },
+            "optional": {
+                "auto_download": ("BOOLEAN", {"default": True}),
+            },
         }
 
     RETURN_TYPES = ("IMAGE",)
@@ -366,6 +383,7 @@ class MTB_DeepBump:
         color_to_normals_overlap="SMALL",
         normals_to_curvature_blur_radius="SMALL",
         normals_to_height_seamless=True,
+        auto_download=False,
     ):
         images = tensor2pil(image)
         out_images = []
@@ -380,7 +398,10 @@ class MTB_DeepBump:
             # Apply processing
             if mode == "Color to Normals":
                 out_img = color_to_normals(
-                    in_img, color_to_normals_overlap, None
+                    in_img,
+                    color_to_normals_overlap,
+                    None,
+                    auto_download=auto_download,
                 )
             if mode == "Normals to Curvature":
                 out_img = normals_to_curvature(

nodes/generate.py

@@ -1,4 +1,8 @@
-from PIL import Image
+import io
+
+import requests
+import torch
+from PIL import Image, ImageDraw, ImageFont
 
 from ..log import log
 from ..utils import comfy_dir, font_path, pil2tensor
@@ -81,10 +85,6 @@ class MTB_UnsplashImage:
     CATEGORY = "mtb/generate"
 
     def do_unsplash_image(self, width, height, random_seed, keyword=None):
-        import io
-
-        import requests
-
         base_url = "https://source.unsplash.com/random/"
 
         if width and height:
@@ -213,14 +213,90 @@ by default it fallsback to a default font.
                     "INT",
                     {"default": 100, "min": 1, "max": 100, "step": 1},
                 ),
-            }
+            },
+            "optional": {
+                "whisper_chunks": ("WHISPER_CHUNKS",),
+                "fps": (
+                    "INT",
+                    {"default": 24, "min": 1, "max": 60, "step": 1},
+                ),
+                "fade_duration": (
+                    "FLOAT",
+                    {"default": 0.5, "min": 0.0, "max": 5.0, "step": 0.1},
+                ),
+            },
         }
 
     RETURN_TYPES = ("IMAGE",)
-    RETURN_NAMES = ("image",)
     FUNCTION = "text_to_image"
     CATEGORY = "mtb/generate"
 
+    def create_animation_frames(
+        self,
+        chunks,
+        base_image,
+        font,
+        font_size,
+        color,
+        width,
+        height,
+        fps,
+        fade_duration,
+    ):
+        """Create animation frames from Whisper chunks."""
+        if not chunks or not chunks.get("chunks"):
+            return [base_image]
+
+        frames = []
+        total_duration = chunks["chunks"][-1]["timestamp"][1]
+        frame_count = int(total_duration * fps)
+        fade_frames = int(fade_duration * fps)
+
+        for frame_idx in range(frame_count):
+            time = frame_idx / fps
+            frame = base_image.copy()
+            draw = ImageDraw.Draw(frame)
+
+            active_chunks = []
+            for chunk in chunks["chunks"]:
+                start, end = chunk["timestamp"]
+                if start <= time <= end:
+                    fade_in_alpha = min(
+                        1.0, (time - start) * fps / fade_frames
+                    )
+                    fade_out_alpha = min(1.0, (end - time) * fps / fade_frames)
+                    alpha = min(fade_in_alpha, fade_out_alpha)
+                    active_chunks.append((chunk["text"], alpha))
+
+            y = height // 4
+            for text, alpha in active_chunks:
+                # Create a temporary image for the text with alpha
+                text_img = Image.new("RGBA", (width, height), (0, 0, 0, 0))
+                text_draw = ImageDraw.Draw(text_img)
+
+                text_draw.text(
+                    (width // 2, y),
+                    text,
+                    font=font,
+                    fill=color,
+                    anchor="mm",
+                )
+
+                text_img.putalpha(
+                    Image.fromarray(
+                        (torch.ones((height, width)) * (alpha * 255))
+                        .byte()
+                        .numpy()
+                    )
+                )
+
+                frame = Image.alpha_composite(frame, text_img)
+                y += font_size * 1.5
+
+            frames.append(frame)
+
+        return frames
+
     def text_to_image(
         self,
         text: str,
@@ -238,58 +314,121 @@ by default it fallsback to a default font.
         h_offset=0,
         v_offset=0,
         h_coverage=100,
+        whisper_chunks=None,
+        fps=24,
+        fade_duration=0.5,
     ):
+        """Convert text to image, with optional animation support."""
         import textwrap
 
-        from PIL import Image, ImageDraw, ImageFont
+        from PIL import ImageColor
 
         font_path = self.fonts[font]
-
-        text = (
-            text.encode("ascii", "ignore").decode().strip() if trim else text
-        )
-        # Handle word wrapping
-        if wrap:
-            wrap_width = (((width / 100) * h_coverage) / font_size) * 2
-            lines = textwrap.wrap(text, width=wrap_width)
-        else:
-            lines = [text]
         font = ImageFont.truetype(font_path, size=font_size)
-        log.debug(f"Lines: {lines}")
-        img = Image.new("RGBA", (width, height), background)
-        draw = ImageDraw.Draw(img)
 
-        line_height_px = line_height * font_size
+        try:
+            if isinstance(color, str):
+                color = ImageColor.getrgb(color)
+            if isinstance(background, str):
+                background = ImageColor.getrgb(background)
 
-        # Vertical alignment
-        if v_align == "top":
-            y_text = v_offset
-        elif v_align == "center":
-            y_text = ((height - (line_height_px * len(lines))) // 2) + v_offset
-        else:  # bottom
-            y_text = (height - (line_height_px * len(lines))) - v_offset
+            if len(color) == 3:
+                color = color + (255,)
+            if len(background) == 3:
+                background = background + (255,)
+        except ValueError as e:
+            log.error(f"Color parsing error: {e}")
+            color = (255, 255, 255, 255)
+            background = (0, 0, 0, 255)
 
-        def get_width(line):
-            if hasattr(font, "getsize"):
-                return font.getsize(line)[0]
+        def render_text(text_to_render, alpha=None):
+            if trim:
+                text_to_render = text_to_render.strip()
+            if wrap:
+                wrap_width = (((width / 100) * h_coverage) / font_size) * 2
+                lines = textwrap.wrap(text_to_render, width=wrap_width)
             else:
-                return font.getlength(line)
+                lines = [text_to_render]
 
-        # Draw each line of text
-        for line in lines:
-            line_width = get_width(line)
-            # Horizontal alignment
-            if h_align == "left":
-                x_text = h_offset
-            elif h_align == "center":
-                x_text = ((width - line_width) // 2) + h_offset
-            else:  # right
-                x_text = (width - line_width) - h_offset
+            img = Image.new("RGBA", (width, height), (0, 0, 0, 0))
+            draw = ImageDraw.Draw(img)
 
-            draw.text((x_text, y_text), line, fill=color, font=font)
-            y_text += line_height_px
+            line_height_px = line_height * font_size
 
-        return (pil2tensor(img),)
+            if v_align == "top":
+                y_text = v_offset
+            elif v_align == "center":
+                y_text = (
+                    (height - (line_height_px * len(lines))) // 2
+                ) + v_offset
+            else:
+                y_text = (height - (line_height_px * len(lines))) - v_offset
+
+            def get_width(line):
+                if hasattr(font, "getsize"):
+                    return font.getsize(line)[0]
+                else:
+                    return font.getlength(line)
+
+            for line in lines:
+                line_width = get_width(line)
+                if h_align == "left":
+                    x_text = h_offset
+                elif h_align == "center":
+                    x_text = ((width - line_width) // 2) + h_offset
+                else:
+                    x_text = (width - line_width) - h_offset
+
+                text_color = color
+                if alpha is not None:
+                    text_color = tuple(
+                        list(color[:3]) + [int(alpha * color[3])]
+                    )
+
+                draw.text((x_text, y_text), line, fill=text_color, font=font)
+                y_text += line_height_px
+
+            return img
+
+        base_img = Image.new("RGBA", (width, height), background)
+
+        if whisper_chunks and whisper_chunks.get("chunks"):
+            frames = []
+            total_duration = whisper_chunks["chunks"][-1]["timestamp"][1]
+            frame_count = int(total_duration * fps)
+            fade_frames = int(fade_duration * fps)
+
+            for frame_idx in range(frame_count):
+                time = frame_idx / fps
+                frame = base_img.copy()
+
+                active_chunks = []
+                for chunk in whisper_chunks["chunks"]:
+                    start, end = chunk["timestamp"]
+                    if start <= time <= end:
+                        fade_in_alpha = min(
+                            1.0, (time - start) * fps / fade_frames
+                        )
+                        fade_out_alpha = min(
+                            1.0, (end - time) * fps / fade_frames
+                        )
+                        alpha = min(fade_in_alpha, fade_out_alpha)
+                        active_chunks.append((chunk["text"], alpha))
+
+                for chunk_text, alpha in active_chunks:
+                    chunk_img = render_text(
+                        chunk_text.encode("ascii", "ignore").decode(), alpha
+                    )
+                    frame = Image.alpha_composite(frame, chunk_img)
+
+                frames.append(frame)
+
+            frame_tensors = [pil2tensor(frame) for frame in frames]
+            return (torch.cat(frame_tensors, dim=0),)
+        else:
+            text_img = render_text(text)
+            result = Image.alpha_composite(base_img, text_img)
+            return (pil2tensor(result),)
 
 
 __nodes__ = [

nodes/graph_utils.py

@@ -1,5 +1,6 @@
 import io
 import json
+import re
 import urllib.parse
 import urllib.request
 from math import pi
@@ -44,14 +45,22 @@ class MTB_ToDevice:
         if torch.backends.mps.is_available():
             devices.append("mps")
         if torch.cuda.is_available():
+            devices.append("cuda:0")
+            for i in range(1, torch.cuda.device_count()):
+                devices.append(f"cuda:{i}")
             devices.append("cuda")
-            for i in range(torch.cuda.device_count()):
-                devices.append(f"cuda{i}")
 
         return {
             "required": {
                 "ignore_errors": ("BOOLEAN", {"default": False}),
-                "device": (devices, {"default": "cpu"}),
+                "device": (
+                    devices,
+                    {
+                        "default": "cuda"
+                        if torch.cuda.is_available()
+                        else "cpu"
+                    },
+                ),
             },
             "optional": {
                 "image": ("IMAGE",),
@@ -67,20 +76,36 @@ class MTB_ToDevice:
     def to_device(
         self,
         *,
-        ignore_errors=False,
-        device="cuda",
+        ignore_errors: bool = False,
+        device: str = "cuda",
         image: torch.Tensor | None = None,
         mask: torch.Tensor | None = None,
     ):
         if not ignore_errors and image is None and mask is None:
             raise ValueError(
                 "You must either provide an image or a mask,"
-                " use ignore_error to passthrough"
+                + " use ignore_error to passthrough"
+            )
+        if (
+            device.startswith("cuda")
+            and ":" not in device
+            and device != "cuda"
+        ):
+            device = f"cuda:{device[4:]}"
+
+        try:
+            if image is not None:
+                image = image.to(device)
+            if mask is not None:
+                mask = mask.to(device)
+        except RuntimeError as e:
+            if not ignore_errors:
+                raise RuntimeError(
+                    f"Failed to move tensor to device {device}: {str(e)}"
+                ) from e
+            log.warning(
+                f"Failed to move tensor to device {device}, ignoring: {str(e)}"
             )
-        if image is not None:
-            image = image.to(device)
-        if mask is not None:
-            mask = mask.to(device)
         return (image, mask)
 
 
@@ -447,7 +472,7 @@ class MTB_AnyToString:
 
 
 class MTB_StringReplace:
-    """Basic string replacement."""
+    """Basic string replacement with regex support."""
 
     @classmethod
     def INPUT_TYPES(cls):
@@ -456,6 +481,7 @@ class MTB_StringReplace:
                 "string": ("STRING", {"forceInput": True}),
                 "old": ("STRING", {"default": ""}),
                 "new": ("STRING", {"default": ""}),
+                "use_regex": ("BOOLEAN", {"default": False}),
             }
         }
 
@@ -463,12 +489,19 @@ class MTB_StringReplace:
     RETURN_TYPES = ("STRING",)
     CATEGORY = "mtb/string"
 
-    def replace_str(self, string: str, old: str, new: str):
+    def replace_str(self, string: str, old: str, new: str, use_regex: bool):
         log.debug(f"Current string: {string}")
         log.debug(f"Find string: {old}")
         log.debug(f"Replace string: {new}")
+        log.debug(f"Use regex: {use_regex}")
 
-        string = string.replace(old, new)
+        if use_regex:
+            try:
+                string = re.sub(old, new, string)
+            except re.error as e:
+                raise ValueError(f"Regex error: {e}") from e
+        else:
+            string = string.replace(old, new)
 
         log.debug(f"New string: {string}")
 
@@ -653,6 +686,187 @@ class MTB_ConcatImages:
         return (concatenated,)
 
 
+class MTB_TensorOps:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "tensor": ("IMAGE",),
+                "operation": (
+                    [
+                        "multiply",
+                        "divide",
+                        "add",
+                        "subtract",
+                        "power",
+                        "clamp",
+                        "abs",
+                        "log",
+                        "exp",
+                        "convert_dtype",
+                        "normalize_range",
+                        "normalize_per_channel",
+                    ],
+                    {"default": "multiply"},
+                ),
+                "value": (
+                    "FLOAT",
+                    {
+                        "default": 1.0,
+                        "min": -1000000.0,
+                        "max": 1000000.0,
+                        "step": 0.01,
+                    },
+                ),
+                "source_min": (
+                    "FLOAT",
+                    {
+                        "default": 0.0,
+                        "min": -1000000.0,
+                        "max": 1000000.0,
+                        "step": 0.01,
+                    },
+                ),
+                "source_max": (
+                    "FLOAT",
+                    {
+                        "default": 1.0,
+                        "min": -1000000.0,
+                        "max": 1000000.0,
+                        "step": 0.01,
+                    },
+                ),
+                "target_min": (
+                    "FLOAT",
+                    {
+                        "default": 0.0,
+                        "min": -1000000.0,
+                        "max": 1000000.0,
+                        "step": 0.01,
+                    },
+                ),
+                "target_max": (
+                    "FLOAT",
+                    {
+                        "default": 16.0,
+                        "min": -1000000.0,
+                        "max": 1000000.0,
+                        "step": 0.01,
+                    },
+                ),
+                "dtype": (
+                    ["uint8", "float32", "float16", "bfloat16"],
+                    {"default": "float32"},
+                ),
+                "use_mean": ("BOOLEAN", {"default": False}),
+            },
+            "optional": {
+                "target_tensor": ("IMAGE",),
+            },
+        }
+
+    RETURN_TYPES = ("IMAGE",)
+    FUNCTION = "apply"
+    CATEGORY = "mtb/tensor_ops"
+
+    def apply(
+        self,
+        tensor,
+        operation="multiply",
+        value=1.0,
+        source_min=0.0,
+        source_max=1.0,
+        target_min=0.0,
+        target_max=1.0,
+        dtype="float32",
+        use_mean=False,
+        target_tensor=None,
+    ):
+        log.debug(
+            f"Input tensor stats: shape={tensor.shape}, dtype={tensor.dtype}, range=[{tensor.min().item():.6f}, {tensor.max().item():.6f}]"
+        )
+        if operation == "normalize_per_channel":
+            if target_tensor is None:
+                raise ValueError(
+                    "Target tensor required for per-channel normalization"
+                )
+
+            result = tensor.clone()
+            for c in range(tensor.shape[-1]):
+                if use_mean:
+                    source_mean = tensor[..., c].mean()
+                    target_mean = target_tensor[..., c].mean()
+                    scale = target_mean / source_mean
+                    result[..., c] = tensor[..., c] * scale
+                else:
+                    source_min = tensor[..., c].min()
+                    source_max = tensor[..., c].max()
+                    target_min = target_tensor[..., c].min()
+                    target_max = target_tensor[..., c].max()
+
+                    normalized = (tensor[..., c] - source_min) / (
+                        source_max - source_min
+                    )
+                    result[..., c] = (
+                        normalized * (target_max - target_min) + target_min
+                    )
+
+                log.debug(
+                    f"Channel {c} - Scale: source=[{source_min:.6f}, {source_max:.6f}], target=[{target_min:.6f}, {target_max:.6f}]"
+                )
+
+        elif operation == "normalize_range":
+            if target_tensor is not None:
+                target_min = target_tensor.min().item()
+                target_max = target_tensor.max().item()
+                log.debug(
+                    f"Using target tensor range: [{target_min:.6f}, {target_max:.6f}]"
+                )
+
+            normalized = (tensor - source_min) / (source_max - source_min)
+            result = normalized * (target_max - target_min) + target_min
+        elif operation == "convert_dtype":
+            if dtype == "float32":
+                result = tensor.float()
+            elif dtype == "float16":
+                result = tensor.half()
+            elif dtype == "bfloat16":
+                result = tensor.bfloat16()
+
+        else:
+            result = tensor
+            if operation == "multiply":
+                result = tensor * value
+            elif operation == "divide":
+                result = tensor / value if value != 0 else tensor
+            elif operation == "add":
+                result = tensor + value
+            elif operation == "subtract":
+                result = tensor - value
+            elif operation == "power":
+                result = torch.pow(tensor, value)
+            elif operation == "clamp":
+                if target_tensor is not None:
+                    result = torch.clamp(
+                        tensor,
+                        target_tensor.min().item(),
+                        target_tensor.max().item(),
+                    )
+                else:
+                    result = torch.clamp(tensor, source_min, source_max)
+            elif operation == "abs":
+                result = torch.abs(tensor)
+            elif operation == "log":
+                result = torch.log(tensor.clamp(min=1e-10))
+            elif operation == "exp":
+                result = torch.exp(tensor)
+
+        log.debug(
+            f"Output tensor stats: shape={result.shape}, dtype={result.dtype}, range=[{result.min().item():.6f}, {result.max().item():.6f}]"
+        )
+        return (result,)
+
+
 __nodes__ = [
     MTB_StringReplace,
     MTB_FitNumber,
@@ -667,4 +881,5 @@ __nodes__ = [
     MTB_FloatsToFloat,
     MTB_FloatToFloats,
     MTB_FloatsToInts,
+    MTB_TensorOps,
 ]

nodes/image_processing.py

@@ -3,11 +3,12 @@ import json
 import math
 import os
 
-import comfy.model_management as model_management
+import comfy.utils
 import folder_paths
 import numpy as np
 import torch
 import torch.nn.functional as F
+from comfy import model_management
 from PIL import Image, ImageOps
 from PIL.PngImagePlugin import PngInfo
 from skimage.filters import gaussian
@@ -74,7 +75,10 @@ class MTB_ExtractCoordinatesFromImage:
     def INPUT_TYPES(cls):
         return {
             "required": {
-                "threshold": ("FLOAT",),
+                "threshold": (
+                    "FLOAT",
+                    {"default": 0.5, "min": 0.0, "max": 1.0, "step": 0.01},
+                ),
                 "max_points": ("INT", {"default": 50, "min": 0}),
             },
             "optional": {"image": ("IMAGE",), "mask": ("MASK",)},
@@ -87,72 +91,124 @@ class MTB_ExtractCoordinatesFromImage:
         image: torch.Tensor | None = None,
         mask: torch.Tensor | None = None,
     ) -> tuple[list[list[tuple[int, int]]], torch.Tensor]:
-        if image is not None:
-            batch_count, height, width, channel_count = image.shape
-            imgs = image
-        else:
-            if mask is None:
-                raise ValueError("Must provide either image or mask")
-            batch_count, height, width = mask.shape
-            channel_count = 1
-            imgs = mask
+        if image is None and mask is None:
+            raise ValueError("Must provide either image or mask")
 
-        if channel_count not in [1, 2, 3, 4]:
-            raise ValueError(f"Incorrect channel count: {channel_count}")
+        if image is not None:
+            batch_count, height, width, _channel_count = image.shape
+            input_device = image.device
+            if mask is not None:
+                if mask.ndim == 2:
+                    mask = mask.unsqueeze(0)
+                if mask.ndim != 3:
+                    raise ValueError(
+                        f"Mask has unexpected ndim: {mask.ndim}. Expected 2 or 3."
+                    )
+
+                b_mask, h_mask, w_mask = mask.shape
+                if not (h_mask == height and w_mask == width):
+                    raise ValueError(
+                        f"Image dimensions ({height}x{width}) and mask dimensions ({h_mask}x{w_mask}) are spatially incompatible."
+                    )
+                if b_mask == 1 and batch_count > 1:
+                    mask = mask.expand(batch_count, height, width)
+
+                elif b_mask != batch_count:
+                    raise ValueError(
+                        f"Image batch size ({batch_count}) and mask batch size ({b_mask}) are incompatible and mask cannot be broadcast."
+                    )
+        else:
+            if mask.ndim == 2:
+                mask = mask.unsqueeze(0)
+
+            if mask.ndim != 3:
+                raise ValueError(
+                    f"Mask has unexpected ndim: {mask.ndim} when image is not provided. Expected 2 or 3."
+                )
+
+            batch_count, height, width = mask.shape
+            input_device = mask.device
 
         all_points: list[list[tuple[int, int]]] = []
         debug_images = torch.zeros(
             (batch_count, height, width, 3),
             dtype=torch.uint8,
-            device=imgs.device,
+            device=input_device,
         )
 
-        for i, img in enumerate(imgs):
-            if channel_count == 1:
-                alpha_channel = img if len(img.shape) == 2 else img[:, :, 0]
-            elif channel_count == 2:
-                alpha_channel = img[:, :, 1]
-            elif channel_count == 4:
-                alpha_channel = img[:, :, 3]
+        points_tensor = torch.tensor(
+            [255, 255, 255], dtype=torch.uint8, device=input_device
+        )
+
+        for i in range(batch_count):
+            value_threshold: torch.Tensor
+            if image is not None:
+                img_slice = image[i]
+                img_channels = img_slice.shape[2]
+                if img_channels == 1 or img_channels == 2:
+                    value_threshold = img_slice[:, :, 0]
+                elif img_channels == 3 or img_channels == 4:
+                    value_threshold = img_slice[:, :, :3].max(dim=2)[0]
+                else:
+                    raise ValueError(
+                        f"Unsupported image channel count: {img_channels} for image at batch index {i}"
+                    )
             else:
-                # get intensity
-                alpha_channel = img[:, :, :3].max(dim=2)[0]
+                mask_slice = mask[i]
+                value_threshold = mask_slice
 
-            points = (alpha_channel > threshold).nonzero(as_tuple=False)
+            condition = value_threshold > threshold
+            if image is not None and mask is not None:
+                mask_slice = mask[i]
+                mask_active_condition = mask_slice > 0.0
+                condition = condition & mask_active_condition
 
-            if len(points) > max_points:
-                indices = torch.randperm(points.size(0), device=img.device)[
-                    :max_points
-                ]
-                points = points[indices]
+            points_yx = condition.nonzero(as_tuple=False)
 
-            points = [(int(y.item()), int(x.item())) for x, y in points]
-            all_points.append(points)
+            if points_yx.size(0) > max_points:
+                # shuffle and pick max_points randomly
+                indices = torch.randperm(
+                    points_yx.size(0), device=input_device
+                )[:max_points]
+                points_yx = points_yx[indices]
+            elif max_points == 0:
+                points_yx = torch.empty(
+                    (0, 2), dtype=torch.long, device=input_device
+                )
 
-            for x, y in points:
-                self._draw_circle(debug_images[i], (x, y), 5)
+            current_points = [
+                (int(p[1].item()), int(p[0].item())) for p in points_yx
+            ]
+            all_points.append(current_points)
+            for x_coord, y_coord in current_points:
+                self._draw_circle(
+                    debug_images[i],
+                    (x_coord, y_coord),
+                    radius=5,
+                    color_tensor=points_tensor,
+                )
 
         return (all_points, debug_images)
 
     @staticmethod
     def _draw_circle(
-        image: torch.Tensor, center: tuple[int, int], radius: int
+        image: torch.Tensor,
+        center: tuple[int, int],
+        radius: int,
+        color_tensor: torch.Tensor,
     ):
         """Draw a 5px circle on the image."""
         x0, y0 = center
-        for x in range(-radius, radius + 1):
-            for y in range(-radius, radius + 1):
-                in_radius = x**2 + y**2 <= radius**2
-                in_bounds = (
-                    0 <= x0 + x < image.shape[1]
-                    and 0 <= y0 + y < image.shape[0]
-                )
-                if in_radius and in_bounds:
-                    image[y0 + y, x0 + x] = torch.tensor(
-                        [255, 255, 255],
-                        dtype=torch.uint8,
-                        device=image.device,
-                    )
+        h, w, _ = image.shape
+        min_x_bbox = max(0, x0 - radius)
+        max_x_bbox = min(w - 1, x0 + radius)
+        min_y_bbox = max(0, y0 - radius)
+        max_y_bbox = min(h - 1, y0 + radius)
+
+        for py in range(min_y_bbox, max_y_bbox + 1):
+            for px in range(min_x_bbox, max_x_bbox + 1):
+                if (px - x0) ** 2 + (py - y0) ** 2 <= radius**2:
+                    image[py, px] = color_tensor
 
 
 class MTB_ColorCorrectGPU:
@@ -543,7 +599,7 @@ class MTB_ColorCorrect:
         adjusted = self.hsv_adjustment(adjusted, hue, saturation, value)
 
         if clamp:
-            adjusted = torch.clamp(image, 0.0, 1.0)
+            adjusted = torch.clamp(adjusted, 0.0, 1.0)
 
         result = (
             adjusted
@@ -702,7 +758,6 @@ class MTB_Blur:
                 )
                 blurred_images.append(blurred)
 
-            image_np = np.array(blurred_images)
         else:
             for i in range(image.size(0)):
                 blurred = gaussian(
@@ -710,8 +765,7 @@ class MTB_Blur:
                 )
                 blurred_images.append(blurred)
 
-            image_np = np.array(blurred_images)
-        return (np2tensor(image_np).squeeze(0),)
+        return (np2tensor(blurred_images),)
 
 
 class MTB_Sharpen:

nodes/image_utils.py

@@ -1,4 +1,11 @@
+import json
+import os
+
+import numpy as np
 import torch
+from comfy.cli_args import args
+from PIL import Image
+from PIL.PngImagePlugin import PngInfo
 
 from ..log import log
 
@@ -8,13 +15,21 @@ class MTB_StackImages:
 
     @classmethod
     def INPUT_TYPES(cls):
-        return {"required": {"vertical": ("BOOLEAN", {"default": False})}}
+        return {
+            "required": {"vertical": ("BOOLEAN", {"default": False})},
+            "optional": {
+                "match_method": (
+                    ["error", "smallest", "largest"],
+                    {"default": "error"},
+                )
+            },
+        }
 
     RETURN_TYPES = ("IMAGE",)
     FUNCTION = "stack"
     CATEGORY = "mtb/image utils"
 
-    def stack(self, vertical, **kwargs):
+    def stack(self, vertical, match_method="error", **kwargs):
         if not kwargs:
             raise ValueError("At least one tensor must be provided.")
 
@@ -24,31 +39,62 @@ class MTB_StackImages:
             f"{'vertically' if vertical else 'horizontally'}"
         )
 
+        target_device = tensors[0].device
+
         normalized_tensors = [
-            self.normalize_to_rgba(tensor) for tensor in tensors
+            self.normalize_to_rgba(tensor.to(target_device))
+            for tensor in tensors
         ]
+
         max_batch_size = max(tensor.shape[0] for tensor in normalized_tensors)
         normalized_tensors = [
             self.duplicate_frames(tensor, max_batch_size)
             for tensor in normalized_tensors
         ]
-
-        if vertical:
-            width = normalized_tensors[0].shape[2]
-            if any(tensor.shape[2] != width for tensor in normalized_tensors):
-                raise ValueError(
-                    "All tensors must have the same width "
-                    "for vertical stacking."
+        if match_method != "error":
+            if vertical:
+                # match widths
+                widths = [tensor.shape[2] for tensor in normalized_tensors]
+                target_width = (
+                    min(widths) if match_method == "smallest" else max(widths)
                 )
-            dim = 1
+                normalized_tensors = [
+                    self.resize_tensor(tensor, width=target_width)
+                    for tensor in normalized_tensors
+                ]
+            else:
+                # match heights
+                heights = [tensor.shape[1] for tensor in normalized_tensors]
+                target_height = (
+                    min(heights)
+                    if match_method == "smallest"
+                    else max(heights)
+                )
+                normalized_tensors = [
+                    self.resize_tensor(tensor, height=target_height)
+                    for tensor in normalized_tensors
+                ]
         else:
-            height = normalized_tensors[0].shape[1]
-            if any(tensor.shape[1] != height for tensor in normalized_tensors):
-                raise ValueError(
-                    "All tensors must have the same height "
-                    "for horizontal stacking."
-                )
-            dim = 2
+            if vertical:
+                width = normalized_tensors[0].shape[2]
+                if any(
+                    tensor.shape[2] != width for tensor in normalized_tensors
+                ):
+                    raise ValueError(
+                        "All tensors must have the same width "
+                        "for vertical stacking."
+                    )
+            else:
+                height = normalized_tensors[0].shape[1]
+                if any(
+                    tensor.shape[1] != height for tensor in normalized_tensors
+                ):
+                    raise ValueError(
+                        "All tensors must have the same height "
+                        "for horizontal stacking."
+                    )
+
+        dim = 1 if vertical else 2
 
         stacked_tensor = torch.cat(normalized_tensors, dim=dim)
 
@@ -64,7 +110,7 @@ class MTB_StackImages:
         elif channels == 3:
             alpha_channel = torch.ones(
                 tensor.shape[:-1] + (1,), device=tensor.device
-            )  # Add an alpha channel
+            )
             return torch.cat((tensor, alpha_channel), dim=-1)
         else:
             raise ValueError(
@@ -87,6 +133,30 @@ class MTB_StackImages:
         else:
             return tensor
 
+    def resize_tensor(self, tensor, width=None, height=None):
+        """Resize tensor to specified width or height while maintaining aspect ratio."""
+        current_height, current_width = tensor.shape[1:3]
+
+        if width is not None and width != current_width:
+            scale_factor = width / current_width
+            new_height = int(current_height * scale_factor)
+            new_width = width
+        elif height is not None and height != current_height:
+            scale_factor = height / current_height
+            new_width = int(current_width * scale_factor)
+            new_height = height
+        else:
+            return tensor
+
+        resized = torch.nn.functional.interpolate(
+            tensor.permute(0, 3, 1, 2),
+            size=(new_height, new_width),
+            mode="bilinear",
+            align_corners=False,
+        )
+
+        return resized.permute(0, 2, 3, 1)
+
 
 class MTB_PickFromBatch:
     """Pick a specific number of images from a batch.
@@ -101,30 +171,117 @@ class MTB_PickFromBatch:
                 "image": ("IMAGE",),
                 "from_direction": (["end", "start"], {"default": "start"}),
                 "count": ("INT", {"default": 1}),
-            }
+            },
+            "optional": {
+                "mask": ("MASK",),
+            },
         }
 
-    RETURN_TYPES = ("IMAGE",)
+    RETURN_TYPES = ("IMAGE", "MASK")
     FUNCTION = "pick_from_batch"
     CATEGORY = "mtb/image utils"
 
-    def pick_from_batch(self, image, from_direction, count):
+    def pick_from_batch(self, image, from_direction, count, mask=None):
         batch_size = image.size(0)
 
         # Limit count to the available number of images in the batch
         count = min(count, batch_size)
-        if count < batch_size:
-            log.warning(
-                f"Requested {count} images, "
-                f"but only {batch_size} are available."
-            )
+
+        selected_masks = None
 
         if from_direction == "end":
             selected_tensors = image[-count:]
+            if mask is not None:
+                selected_masks = mask[-count:]
         else:
             selected_tensors = image[:count]
+            if mask is not None:
+                selected_masks = mask[:count]
 
-        return (selected_tensors,)
+        return (selected_tensors, selected_masks)
 
 
-__nodes__ = [MTB_StackImages, MTB_PickFromBatch]
+import folder_paths
+
+
+class MTB_SaveImage:
+    def __init__(self):
+        self.output_dir = folder_paths.get_output_directory()
+        self.type = "output"
+        self.prefix_append = ""
+        self.compress_level = 4
+
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "images": ("IMAGE", {"tooltip": "The images to save."}),
+                "filename_prefix": (
+                    "STRING",
+                    {
+                        "default": "ComfyUI",
+                        "tooltip": "The prefix for the file to save. This may include formatting information such as %date:yyyy-MM-dd% or %Empty Latent Image.width% to include values from nodes.",
+                    },
+                ),
+            },
+            "hidden": {"prompt": "PROMPT", "extra_pnginfo": "EXTRA_PNGINFO"},
+        }
+
+    RETURN_TYPES = ("IMAGE",)
+    FUNCTION = "save_images"
+
+    # OUTPUT_NODE = True
+
+    CATEGORY = "mtb/image utils"
+    DESCRIPTION = """Saves the input images to your ComfyUI output directory.
+    This behaves exactly like the native SaveImage node but isn't an output node.
+    The reason I made this is to allow 'inlining' image save in loops for instance,
+    using the native node there wouldn't run for each iteration of the loop."""
+
+    def save_images(
+        self,
+        images,
+        filename_prefix="ComfyUI",
+        prompt=None,
+        extra_pnginfo=None,
+    ):
+        filename_prefix += self.prefix_append
+        full_output_folder, filename, counter, subfolder, filename_prefix = (
+            folder_paths.get_save_image_path(
+                filename_prefix,
+                self.output_dir,
+                images[0].shape[1],
+                images[0].shape[0],
+            )
+        )
+        results = list()
+        for batch_number, image in enumerate(images):
+            i = 255.0 * image.cpu().numpy()
+            img = Image.fromarray(np.clip(i, 0, 255).astype(np.uint8))
+            metadata = None
+            if not args.disable_metadata:
+                metadata = PngInfo()
+                if prompt is not None:
+                    metadata.add_text("prompt", json.dumps(prompt))
+                if extra_pnginfo is not None:
+                    for x in extra_pnginfo:
+                        metadata.add_text(x, json.dumps(extra_pnginfo[x]))
+
+            filename_with_batch_num = filename.replace(
+                "%batch_num%", str(batch_number)
+            )
+            file = f"{filename_with_batch_num}_{counter:05}_.png"
+            img.save(
+                os.path.join(full_output_folder, file),
+                pnginfo=metadata,
+                compress_level=self.compress_level,
+            )
+            results.append(
+                {"filename": file, "subfolder": subfolder, "type": self.type}
+            )
+            counter += 1
+
+        return {"ui": {"images": results}, "result": (images,)}
+
+
+__nodes__ = [MTB_StackImages, MTB_PickFromBatch, MTB_SaveImage]

nodes/ltx.py

@@ -42,6 +42,9 @@ class ImageH264Compression:
     DESCRIPTION = """
 **Encodes the input with h264 compression using a configurable CRF**.
 
+> [!IMPORTANT]
+> This node is not really needed with the latest version of LTXVideo.
+
 > [!NOTE]
 > This was recommended by the creators of LTX over banodoco's discord.
 
@@ -151,6 +154,7 @@ class ImageH264Compression:
         output_images = torch.stack(output_images).to(image.device)
         return (output_images,)
 
+
 # fmt: off
 __nodes__ = [
     ImageH264Compression

nodes/transform.py

@@ -45,6 +45,34 @@ class MTB_TransformImage:
                 ),
                 "constant_color": ("COLOR", {"default": "#000000"}),
             },
+            "optional": {
+                "filter_type": (
+                    [
+                        "nearest",
+                        "box",
+                        "bilinear",
+                        "hamming",
+                        "bicubic",
+                        "lanczos",
+                    ],
+                    {"default": "bilinear"},
+                ),
+                "stretch_x": (
+                    "FLOAT",
+                    {"default": 1.0, "min": 0.001, "max": 10.0, "step": 0.01},
+                ),
+                "stretch_y": (
+                    "FLOAT",
+                    {"default": 1.0, "min": 0.001, "max": 10.0, "step": 0.01},
+                ),
+                "use_normalized": (
+                    "BOOLEAN",
+                    {
+                        "default": False,
+                        "tooltip": "If true, transform values are scaled to image dimensions.",
+                    },
+                ),
+            },
         }
 
     FUNCTION = "transform"
@@ -61,21 +89,36 @@ class MTB_TransformImage:
         shear: float,
         border_handling="edge",
         constant_color=None,
+        filter_type="nearest",
+        stretch_x=1.0,
+        stretch_y=1.0,
+        use_normalized: bool = False,
     ):
+        filter_map = {
+            "nearest": Image.NEAREST,
+            "box": Image.BOX,
+            "bilinear": Image.BILINEAR,
+            "hamming": Image.HAMMING,
+            "bicubic": Image.BICUBIC,
+            "lanczos": Image.LANCZOS,
+        }
+        resampling_filter = filter_map[filter_type]
+
+        _, frame_height, frame_width, _ = image.size()
+        if use_normalized:
+            x = float(x) * frame_width
+            y = float(y) * frame_height
         x = int(x)
         y = int(y)
         angle = int(angle)
 
         log.debug(
-            f"Zoom: {zoom} | x: {x}, y: {y}, angle: {angle}, shear: {shear}"
+            f"Zoom: {zoom} | x: {x}, y: {y}, angle: {angle}, shear: {shear} | stretch_x: {stretch_x}, stretch_y: {stretch_y}"
         )
 
         if image.size(0) == 0:
             return (torch.zeros(0),)
         transformed_images = []
-        frames_count, frame_height, frame_width, frame_channel_count = (
-            image.size()
-        )
 
         new_height, new_width = (
             int(frame_height * zoom),
@@ -106,18 +149,55 @@ class MTB_TransformImage:
 
         for img in tensor2pil(image):
             img = TF.pad(
-                img,  # transformed_frame,
+                img,
                 padding=padding,
                 padding_mode=border_handling,
                 fill=constant_color or 0,
             )
 
-            img = cast(
-                Image.Image,
-                TF.affine(
-                    img, angle=angle, scale=zoom, translate=[x, y], shear=shear
-                ),
-            )
+            if stretch_x != 1.0 or stretch_y != 1.0:
+                img = cast(
+                    Image.Image,
+                    TF.affine(
+                        img,
+                        angle=angle,
+                        scale=zoom,
+                        translate=[x, y],
+                        shear=shear,
+                        interpolation=resampling_filter,
+                    ),
+                )
+
+                width, height = img.size
+                center = (width // 2, height // 2)
+
+                stretch_x_factor = 1.0 / stretch_x
+                stretch_y_factor = 1.0 / stretch_y
+
+                matrix = [
+                    stretch_x_factor,
+                    0,
+                    center[0] - center[0] * stretch_x_factor,
+                    0,
+                    stretch_y_factor,
+                    center[1] - center[1] * stretch_y_factor,
+                ]
+
+                img = img.transform(
+                    img.size, Image.AFFINE, matrix, resampling_filter
+                )
+            else:
+                img = cast(
+                    Image.Image,
+                    TF.affine(
+                        img,
+                        angle=angle,
+                        scale=zoom,
+                        translate=[x, y],
+                        shear=shear,
+                        interpolation=resampling_filter,
+                    ),
+                )
 
             left = abs(padding[0])
             upper = abs(padding[1])

pyproject.toml

@@ -4,9 +4,9 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "comfy-mtb"
-version = "0.2.0"
+version = "0.5.0"
 description = "Animation oriented nodes pack for ComfyUI."
-license = "MIT"
+license = { text = "MIT" }
 readme = "README.md"
 # repository = ""
 # url = "https://github.com/melMass/comfy_mtb"
@@ -38,6 +38,7 @@ optional-dependencies = { mel = [
 ], dev = [
   "black[jupyter]",
   "codespell",
+  "marimo",
   "mypy",
   "pre-commit",
   "pytest",
@@ -62,7 +63,7 @@ DisplayName = "comfy-mtb"
 Icon = "https://avatars.githubusercontent.com/u/7041726?v=4"
 
 [tool.bumpversion]
-current_version = "0.2.0"
+current_version = "0.5.0"
 parse = "(?P<major>\\d+)\\.(?P<minor>\\d+)\\.(?P<patch>\\d+)"
 serialize = ["{major}.{minor}.{patch}"]
 search = "{current_version}"
@@ -149,6 +150,9 @@ show_contexts = true
 
 [tool.ruff]
 line-length = 79
+extend-exclude = ["./docs/conf.py", "notebooks", "stubs"]
+
+[tool.ruff.lint]
 select = ["A", "B", "C", "D", "E", "F", "FBT", "I", "N", "S", "SIM", "UP", "W"]
 # NOTE:
 # D102 - undocumented-public-method (noisy)
@@ -156,16 +160,14 @@ select = ["A", "B", "C", "D", "E", "F", "FBT", "I", "N", "S", "SIM", "UP", "W"]
 # D100 - undocumented-public-module (noisy)
 # N802 - invalid-function-name (forced by comfy's arch)
 ignore = ["D103", "D102", "D100", "N802"]
-# exclude auto generated file
-extend-exclude = ["./docs/conf.py"]
 
-[tool.ruff.per-file-ignores]
+[tool.ruff.lint.per-file-ignores]
 # imported but unused
 "__init__.py" = ["F401"]
 # use of assert detected
 "tests/*" = ["S101"]
 
-[tool.ruff.pydocstyle]
+[tool.ruff.lint.pydocstyle]
 convention = "numpy"
 
 [tool.mypy]

requirements.txt

@@ -9,3 +9,4 @@ rich_argparse
 matplotlib
 pillow
 cachetools
+transformers

types/typedefs.js

@@ -16,3 +16,50 @@
  * @typedef {import("./shared.d.ts").INodeOutputSlot} INodeOutputSlot
  */
 
+/**
+ * @typedef {Object} ResultItem
+ * @property {string} [filename] - The filename of the item.
+ * @property {string} [subfolder] - The subfolder of the item.
+ * @property {string} [type] - The type of the item.
+ */
+
+/**
+ * @typedef {Object} Outputs
+ * @property {ResultItem[]} [audio] - Audio result items.
+ * @property {ResultItem[]} [images] - Image result items.
+ * @property {ResultItem[]} [animated] - Animated result items.
+ */
+
+/**
+ * @typedef {Record<string, Outputs>} TaskOutput
+ * - A record mapping Node IDs to their Outputs.
+ */
+
+/**
+ * @typedef {Array} TaskPrompt
+ * @property {QueueIndex} [0] - The queue index.
+ * @property {PromptId} [1] - The unique prompt ID.
+ * @property {PromptInputs} [2] - The prompt inputs.
+ * @property {ExtraData} [3] - Extra data.
+ * @property {OutputsToExecute} [4] - The outputs to execute.
+ */
+
+/**
+ * @typedef {Object} HistoryTaskItem
+ * @property {'History'} taskType - The type of task.
+ * @property {TaskPrompt} prompt - The task prompt.
+ * @property {Status} [status] - The status of the task.
+ * @property {TaskOutput} outputs - The task outputs.
+ * @property {TaskMeta} [meta] - Optional task metadata.
+ */
+
+/**
+ * @typedef {Object} ExecInfo
+ * @property {number} queue_remaining - The number of items remaining in the queue.
+ */
+
+/**
+ * @typedef {Object} StatusWsMessageStatus
+ * @property {ExecInfo} exec_info - Execution information.
+ */
+

utils.py

@@ -2,6 +2,7 @@ import contextlib
 import functools
 import importlib
 import math
+import operator
 import os
 import shlex
 import shutil
@@ -11,9 +12,12 @@ import sys
 import uuid
 from collections.abc import Callable, Sequence
 from enum import Enum
+from functools import reduce
 from pathlib import Path
 from typing import TypeVar
+from urllib.parse import urlparse
 
+import comfy.utils
 import folder_paths
 import numpy as np
 import numpy.typing as npt
@@ -212,6 +216,37 @@ def get_server_info():
 
 
 # region MISC Utilities
+def glob_multiple(
+    path: Path, patterns: list[str], recursive: bool = False
+) -> list[Path]:
+    """Combine multiple glob patterns into a single iterator."""
+    return list(reduce(operator.or_, (set(path.glob(p)) for p in patterns)))
+
+
+def build_glob_patterns(
+    extensions: list[str], recursive: bool = False
+) -> list[str]:
+    """Build glob patterns for given extensions."""
+    prefix = "**/" if recursive else ""
+    return [f"{prefix}*.{ext}" for ext in extensions]
+
+
+class SortMode(Enum):
+    NONE = "none"
+    MODIFIED = "modified"
+    MODIFIED_REVERSE = "modified-reverse"
+    NAME = "name"
+    NAME_REVERSE = "name-reverse"
+
+    @classmethod
+    def from_str(cls, value: str | None) -> "SortMode|None":
+        if not value:
+            return None
+        try:
+            return cls(value.lower())
+        except ValueError:
+            log.warning(f"Sort mode {value} not supported")
+            return None
 
 
 # TODO: use mtb.core directly instead of copying parts here
@@ -465,8 +500,12 @@ here = Path(__file__).parent.absolute()
 # - Construct the absolute path to the ComfyUI directory
 comfy_dir = Path(folder_paths.base_path)
 models_dir = Path(folder_paths.models_dir)
+
+
+# NOTE: these aren't reliable, better call the getters each time
 output_dir = Path(folder_paths.output_directory)
 input_dir = Path(folder_paths.input_directory)
+
 styles_dir = comfy_dir / "styles"
 session_id = str(uuid.uuid4())
 # - Construct the path to the font file
@@ -476,9 +515,10 @@ font_path = here / "data" / "font.ttf"
 extern_root = here / "extern"
 add_path(extern_root)
 
-for pth in extern_root.iterdir():
-    if pth.is_dir():
-        add_path(pth)
+if extern_root.exists():
+    for pth in extern_root.iterdir():
+        if pth.is_dir():
+            add_path(pth)
 
 # - Add the ComfyUI directory and custom nodes path to the sys.path list
 add_path(comfy_dir)
@@ -506,9 +546,7 @@ PIL_FILTER_MAP = {
 # region TENSOR Utilities
 def to_numpy(image: torch.Tensor) -> npt.NDArray[np.uint8]:
     """Converts a tensor to a ndarray with proper scaling and type conversion."""
-    log.debug(f"Converting tensor to numpy array with shape {image.shape}")
     np_array = np.clip(255.0 * image.cpu().numpy(), 0, 255).astype(np.uint8)
-    log.debug(f"Numpy array shape after conversion: {np_array.shape}")
     return np_array
 
 
@@ -820,6 +858,62 @@ def tiles_split(img, tile_size, stride_size):
 
 
 # region MODEL Utilities
+
+
+def download_model(model_url: str, destination: str):
+    if isinstance(model_url, list):
+        for url in model_url:
+            download_model(url, destination)
+        return
+
+    filename = Path(urlparse(model_url).path).name
+
+    if "drive.google.com" in model_url:
+        try:
+            import gdown
+        except ImportError:
+            log.info("Installing gdown")
+            subprocess.check_call(
+                [
+                    sys.executable,
+                    "-m",
+                    "pip",
+                    "install",
+                    "gdown",
+                ]
+            )
+            import gdown
+
+        if "/folders/" in model_url:
+            # download folder
+            try:
+                gdown.download_folder(
+                    model_url, output=destination, resume=True
+                )
+            except TypeError:
+                gdown.download_folder(model_url, output=destination)
+
+            return
+        # download from google drive
+        gdown.download(model_url, destination, quiet=False, resume=True)
+        return True
+    response = requests.get(model_url, stream=True)
+    total_size = int(response.headers.get("content-length", 0))
+
+    destination_path = get_model_path(destination, filename)
+    destination_path.parent.mkdir(exist_ok=True)
+
+    pbar = comfy.utils.ProgressBar(total_size)
+    with open(destination_path, "wb") as file:
+        for data in response.iter_content(chunk_size=4096):
+            file.write(data)
+            pbar.update(len(data))
+
+    log.info(
+        f"Downloaded model from {model_url} to {destination_path}",
+    )
+
+
 def download_antelopev2():
     antelopev2_url = (
         "https://drive.google.com/uc?id=18wEUfMNohBJ4K3Ly5wpTejPfDzp-8fI8"

web/comfy_shared.js

@@ -1,10 +1,9 @@
 /**
+ * @module Shared utilities
  * File: comfy_shared.js
  * Project: comfy_mtb
  * Author: Mel Massadian
- *
  * Copyright (c) 2023-2024 Mel Massadian
- *
  */
 
 // Reference the shared typedefs file
@@ -372,23 +371,37 @@ export function getWidgetType(config) {
  * @param {NodeType} nodeType The nodetype to attach the documentation to
  * @param {str} prefix A prefix added to each dynamic inputs
  * @param {str | [str]} inputType The datatype(s) of those dynamic inputs
- * @param {{link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}?} opts
+ * @param {{separator?:string, start_index?:number, link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}?} [opts] Extra options
  * @returns
  */
-export const setupDynamicConnections = (nodeType, prefix, inputType, opts) => {
+export const setupDynamicConnections = (
+  nodeType,
+  prefix,
+  inputType,
+  opts = undefined,
+) => {
   infoLogger(
     'Setting up dynamic connections for',
     Object.getOwnPropertyDescriptors(nodeType).title.value,
   )
 
-  /** @type {{link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}} */
-  const options = opts || {}
+  /** @type {{separator:string, start_index:number, link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}?} */
+  const options = Object.assign(
+    {
+      separator: '_',
+      start_index: 1,
+    },
+    opts || {},
+  )
   const onNodeCreated = nodeType.prototype.onNodeCreated
   const inputList = typeof inputType === 'object'
 
   nodeType.prototype.onNodeCreated = function () {
     const r = onNodeCreated ? onNodeCreated.apply(this, []) : undefined
-    this.addInput(`${prefix}_1`, inputList ? '*' : inputType)
+    this.addInput(
+      `${prefix}${options.separator}${options.start_index}`,
+      inputList ? '*' : inputType,
+    )
     return r
   }
 
@@ -423,7 +436,7 @@ export const setupDynamicConnections = (nodeType, prefix, inputType, opts) => {
       this,
       slotIndex,
       isConnected,
-      `${prefix}_`,
+      `${prefix}${options.separator}`,
       inputType,
       options,
     )
@@ -439,7 +452,7 @@ export const setupDynamicConnections = (nodeType, prefix, inputType, opts) => {
  * @param {bool} connected - Was this event connecting or disconnecting
  * @param {string} [connectionPrefix] - The common prefix of the dynamic inputs
  * @param {string|[string]} [connectionType] - The type of the dynamic connection
- * @param {{link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}} [opts] - extra options
+ * @param {{start_index?:number, link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}} [opts] - extra options
  */
 export const dynamic_connection = (
   node,
@@ -449,13 +462,18 @@ export const dynamic_connection = (
   connectionType = '*',
   opts = undefined,
 ) => {
-  /* @type {{link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}} [opts] - extra options*/
-  const options = opts || {}
+  /* {{start_index:number, link?:LLink, ioSlot?:INodeInputSlot | INodeOutputSlot}} [opts] - extra options*/
+  const options = Object.assign(
+    {
+      start_index: 1,
+    },
+    opts || {},
+  )
 
-  if (
-    node.inputs.length > 0 &&
-    !node.inputs[index].name.startsWith(connectionPrefix)
-  ) {
+  // function to test if input is a dynamic one
+  const isDynamicInput = (inputName) => inputName.startsWith(connectionPrefix)
+
+  if (node.inputs.length > 0 && !isDynamicInput(node.inputs[index].name)) {
     return
   }
 
@@ -474,7 +492,7 @@ export const dynamic_connection = (
     const to_remove = []
     for (let n = 1; n < node.inputs.length; n++) {
       const element = node.inputs[n]
-      if (!element.link) {
+      if (!element.link && isDynamicInput(element.name)) {
         if (node.widgets) {
           const w = node.widgets.find((w) => w.name === element.name)
           if (w) {
@@ -500,14 +518,25 @@ export const dynamic_connection = (
 
     infoLogger('Cleaning inputs: making it sequential again')
     // make inputs sequential again
+    let prefixed_idx = options.start_index
     for (let i = 0; i < node.inputs.length; i++) {
-      let name = `${connectionPrefix}${i + 1}`
+      let name = ''
+      // rename only prefixed inputs
+      if (isDynamicInput(node.inputs[i].name)) {
+        // prefixed => rename and increase index
+        name = `${connectionPrefix}${prefixed_idx}`
+        prefixed_idx += 1
+      } else {
+        // not prefixed => keep same name
+        name = node.inputs[i].name
+      }
 
       if (nameArray.length > 0) {
         name = i < nameArray.length ? nameArray[i] : name
       }
 
-      node.inputs[i].label = name
+      // preserve label if it exists
+      node.inputs[i].label = node.inputs[i].label || name
       node.inputs[i].name = name
     }
   }
@@ -547,11 +576,16 @@ export const dynamic_connection = (
     if (node.inputs.length === 0) return
     // add an extra input
     if (node.inputs[node.inputs.length - 1].link !== null) {
-      const nextIndex = node.inputs.length
+      // count only the prefixed inputs
+      const nextIndex = node.inputs.reduce(
+        (acc, cur) => (isDynamicInput(cur.name) ? ++acc : acc),
+        0,
+      )
+
       const name =
         nextIndex < nameArray.length
           ? nameArray[nextIndex]
-          : `${connectionPrefix}${nextIndex + 1}`
+          : `${connectionPrefix}${nextIndex + options.start_index}`
 
       infoLogger(`Adding input ${nextIndex + 1} (${name})`)
       node.addInput(name, conType)
@@ -1095,7 +1129,33 @@ export const addDeprecation = (nodeType, reason) => {
 
 // #endregion
 
-// #region API / graph utilities
+// #region Actions API
+export const runAction = async (name, ...args) => {
+  const req = await api.fetchApi('/mtb/actions', {
+    method: 'POST',
+    body: JSON.stringify({
+      name,
+      args,
+    }),
+  })
+
+  const res = await req.json()
+  return res.result
+}
+export const getServerInfo = async () => {
+  const res = await api.fetchApi('/mtb/server-info')
+  return await res.json()
+}
+export const setServerInfo = async (opts) => {
+  await api.fetchApi('/mtb/server-info', {
+    method: 'POST',
+    body: JSON.stringify(opts),
+  })
+}
+
+// #endregion
+
+// #region Authoring API / graph utilities
 export const getAPIInputs = () => {
   const inputs = {}
   let counter = 1
@@ -1148,3 +1208,4 @@ export const getNodes = (skip_unused) => {
   }
   return nodes
 }
+// #endregion

web/debug.js

@@ -11,11 +11,8 @@
 /// <reference path="../types/typedefs.js" />
 
 import { app } from '../../scripts/app.js'
-
 import * as shared from './comfy_shared.js'
-import { MtbWidgets } from './mtb_widgets.js'
-
-// TODO: respect inputs order...
+import * as mtb_ui from './mtb_ui.js'
 
 function escapeHtml(unsafe) {
   return unsafe
@@ -25,6 +22,54 @@ function escapeHtml(unsafe) {
     .replace(/"/g, '&quot;')
     .replace(/'/g, '&#039;')
 }
+
+function createDebugSection(title) {
+  const section = mtb_ui.makeElement('div', {
+    margin: '8px 0',
+    padding: '8px',
+    borderRadius: '4px',
+    backgroundColor: 'rgba(0,0,0,0.2)'
+  })
+
+  const header = mtb_ui.makeElement('h3', {
+    margin: '0 0 8px 0',
+    padding: '4px 0',
+    borderBottom: '1px solid rgba(255,255,255,0.1)',
+    fontSize: '14px',
+    fontWeight: 'bold',
+    color: '#9f9'
+  })
+  header.textContent = title
+  section.appendChild(header)
+
+  return section
+}
+
+function createDebugContent(content, type) {
+  const wrapper = mtb_ui.makeElement('div', {
+    margin: '4px 0'
+  })
+
+  if (type === 'text') {
+    const text = mtb_ui.makeElement('p', {
+      margin: '2px 0',
+      fontFamily: 'monospace',
+      whiteSpace: 'pre-wrap'
+    })
+    text.innerHTML = content
+    wrapper.appendChild(text)
+  } else if (type === 'image') {
+    const img = mtb_ui.makeElement('img', {
+      width: '100%',
+      borderRadius: '2px'
+    })
+    img.src = content
+    wrapper.appendChild(img)
+  }
+
+  return wrapper
+}
+
 app.registerExtension({
   name: 'mtb.Debug',
 
@@ -36,12 +81,10 @@ app.registerExtension({
   async beforeRegisterNodeDef(nodeType, nodeData, app) {
     if (nodeData.name === 'Debug (mtb)') {
       const onNodeCreated = nodeType.prototype.onNodeCreated
-      nodeType.prototype.onNodeCreated = function () {
+      nodeType.prototype.onNodeCreated = function (...args) {
         this.options = {}
-        const r = onNodeCreated
-          ? onNodeCreated.apply(this, arguments)
-          : undefined
-        this.addInput(`anything_1`, '*')
+        const r = onNodeCreated ? onNodeCreated.apply(this, args) : undefined
+        this.addInput('anything_1', '*')
         return r
       }
 
@@ -81,51 +124,82 @@ app.registerExtension({
       }
 
       const onExecuted = nodeType.prototype.onExecuted
-      nodeType.prototype.onExecuted = function (data) {
-        onExecuted?.apply(this, arguments)
-
-        const prefix = 'anything_'
+      nodeType.prototype.onExecuted = function (...args) {
+        onExecuted?.apply(this, args)
+        const [data, ..._rest] = args
 
         if (this.widgets) {
+          let tgt_len = this.widgets.length
           for (let i = 0; i < this.widgets.length; i++) {
-            if (this.widgets[i].name !== 'output_to_console') {
+            if (
+              this.widgets[i].name !== 'output_to_console' &&
+              this.widgets[i].name !== 'as_detailed_types'
+            ) {
+              this.widgets[i].onRemove?.()
               this.widgets[i].onRemoved?.()
+              tgt_len -= 1
             }
           }
-          this.widgets.length = 1
-        }
-        let widgetI = 1
-        // console.log(message)
-        if (data.text) {
-          for (const txt of data.text) {
-            const w = this.addCustomWidget(
-              MtbWidgets.DEBUG_STRING(`${prefix}_${widgetI}`, escapeHtml(txt)),
-            )
-            w.parent = this
-            widgetI++
-          }
-        }
-        if (data.b64_images) {
-          for (const img of data.b64_images) {
-            const w = this.addCustomWidget(
-              MtbWidgets.DEBUG_IMG(`${prefix}_${widgetI}`, img),
-            )
-            w.parent = this
-            widgetI++
-          }
+          this.widgets.length = tgt_len
         }
 
-        // this.setSize(this.computeSize())
+        const inputData = {}
+
+        const uiData = data.ui || data
+
+        if (uiData.items) {
+            uiData.items.forEach(item => {
+                const inputName = item.input
+                if (!inputData[inputName]) {
+                    inputData[inputName] = { text: [], b64_images: [] }
+                }
+                if (item.text) {
+                    inputData[inputName].text.push(...item.text)
+                }
+                if (item.b64_images) {
+                    inputData[inputName].b64_images.push(...item.b64_images)
+                }
+            })
+        }
+
+        let widgetI = 1
+        for (const [inputName, content] of Object.entries(inputData)) {
+          if (content.text.length === 0 && content.b64_images.length === 0) {
+            continue
+          }
+
+          const section = createDebugSection(inputName)
+
+          if (content.text.length > 0) {
+            content.text.forEach(text => {
+              section.appendChild(createDebugContent(text, 'text'))
+            })
+          }
+
+          if (content.b64_images.length > 0) {
+            content.b64_images.forEach(img => {
+              section.appendChild(createDebugContent(img, 'image'))
+            })
+          }
+
+          this.addDOMWidget(
+            `debug_section_${widgetI}`,
+            'CUSTOM',
+            section,
+            {}
+          )
+          widgetI++
+        }
 
         this.onRemoved = function () {
-          // When removing this node we need to remove the input from the DOM
-          for (let y in this.widgets) {
-            if (this.widgets[y].canvas) {
-              this.widgets[y].canvas.remove()
+          for (const widget of this.widgets) {
+            if (widget.canvas) {
+              widget.canvas.remove()
             }
-            shared.cleanupNode(this)
-            this.widgets[y].onRemoved?.()
+            widget.onRemoved?.()
+            widget.onRemove?.()
           }
+          shared.cleanupNode(this)
         }
       }
     }

web/mtb_input_output_sidebar.js

@@ -1,7 +1,9 @@
+/// <reference path="../types/typedefs.js" />
+
 import { app } from '../../scripts/app.js'
 import { api } from '../../scripts/api.js'
 
-// import * as shared from './comfy_shared.js'
+import * as shared from './comfy_shared.js'
 
 import {
   // defineCSSClass,
@@ -15,9 +17,12 @@ import {
 const offset = 0
 let currentWidth = 200
 let currentMode = 'input'
+let subfolder = ''
 let currentSort = 'None'
 
-const IMAGE_NODES = ['LoadImage']
+const IMAGE_NODES = ['LoadImage', 'VHS_LoadImagePath']
+const VIDEO_NODES = ['VHS_LoadVideo']
+const PROCESSED_PROMPT_IDS = new Set()
 
 const updateImage = (node, image) => {
   if (IMAGE_NODES.includes(node.type)) {
@@ -26,26 +31,110 @@ const updateImage = (node, image) => {
       w.value = image
       w.callback()
     }
+  } else if (VIDEO_NODES.includes(node.type)) {
+    const w = node.widgets?.find((w) => w.name === 'video')
+    if (w) {
+      node.updateParameters({ filename: image }, true)
+    }
+  } else {
+    console.warn('No method to update', node.type)
   }
 }
 
-const getImgsFromUrls = (urls, target) => {
+/**
+ * Converts a result item to a request url.
+ * @param {ResultItem} resultItem
+ * @returns {string} - The request URL.
+ */
+const resultItemToQuery = (resultItem) =>
+  [
+    `/mtb/view?filename=${resultItem.filename}`,
+    `width=512`,
+    `type=${resultItem.type}`,
+    `subfolder=${resultItem.subfolder}`,
+    `preview=`,
+  ].join('&')
+
+/**
+ * Retrieves the unique prompt ID from a history task item.
+ * @param {HistoryTaskItem} historyTaskItem
+ * @returns {string} - The prompt ID.
+ */
+const getPromptId = (historyTaskItem) => `${historyTaskItem.prompt[1]}`
+
+/**
+ * Process and return any new/unseen outputs from the most recent history item.
+ * @param {HistoryTaskItem} mostRecentTask - The most recent history task item.
+ * @returns {Object<string, string>} - A map of task outputs URLs.
+ */
+const getNewOutputUrls = (mostRecentTask) => {
+  if (!mostRecentTask) return
+
+  const promptId = getPromptId(mostRecentTask)
+  if (PROCESSED_PROMPT_IDS.has(promptId)) return
+
+  const urls = {}
+  for (const nodeOutputs of Object.values(mostRecentTask.outputs)) {
+    const { images, audio, animated } = nodeOutputs
+    if (images) {
+      const imageOutputs = Object.values(nodeOutputs.images)
+      imageOutputs.forEach(
+        (resultItem) =>
+          (urls[resultItem.filename] = resultItemToQuery(resultItem))
+      )
+    }
+    // Can process `animated` and `audio` outputs here.
+  }
+
+  const foundNewOutputs = Object.keys(urls).length > 0
+  if (!foundNewOutputs) return null
+
+  PROCESSED_PROMPT_IDS.add(promptId)
+  return urls
+}
+
+/** Fetch history and update the grid with any new ouput images. */
+const updateOutputsGrid = async () => {
+  try {
+    const history = await api.getHistory(/** maxSize: */ 1)
+    const mostRcentTask = history.History[0]
+    const newUrls = getNewOutputUrls(mostRcentTask)
+    if (newUrls) {
+      const imgGrid = document.querySelector('.mtb_img_grid')
+      getImgsFromUrls(newUrls, imgGrid, { prepend: true })
+    }
+  } catch (error) {
+    console.error('Error fetching history:', error)
+  }
+}
+
+const getImgsFromUrls = (urls, target, options = { prepend: false }) => {
   const imgs = []
   if (urls === undefined) {
     return imgs
   }
+  const elem = currentMode === 'video' ? 'video' : 'img'
 
   for (const [key, url] of Object.entries(urls)) {
-    const a = makeElement('img')
+    const a = makeElement(elem)
     a.src = url
     a.width = currentWidth
     if (currentMode === 'input') {
       a.onclick = (_e) => {
+        if (subfolder !== '') {
+          app.extensionManager.toast.add({
+            severity: 'warn',
+            summary: 'Subfolder not supported',
+            detail: "The LoadImage node doesn't support subfolders",
+            life: 5000,
+          })
+          return
+        }
         const selected = app.canvas.selected_nodes
         if (selected && Object.keys(selected).length === 0) {
           app.extensionManager.toast.add({
             severity: 'warn',
-            summary: 'No LoadImage node selected!',
+            summary: 'No node selected!',
             detail:
               'For now the only action when clicking images in the sidebar is to set the image on all selected LoadImage nodes.',
             life: 5000,
@@ -54,12 +143,22 @@ const getImgsFromUrls = (urls, target) => {
         }
 
         for (const [_id, node] of Object.entries(app.canvas.selected_nodes)) {
-          updateImage(node, `${key}.png`)
+          updateImage(node, key)
         }
       }
-    } else {
-      a.onclick = (_e) =>
+    } else if (currentMode === 'output') {
+      a.onclick = (_e) => {
         // window.MTB?.notify?.("Output import isn't supported yet...", 5000)
+        if (subfolder !== '') {
+          app.extensionManager.toast.add({
+            severity: 'warn',
+            summary: 'Subfolder not supported',
+            detail: "The LoadImage node doesn't support subfolders",
+            life: 5000,
+          })
+          return
+        }
+
         app.extensionManager.toast.add({
           severity: 'warn',
           summary: 'Outputs not supported',
@@ -67,28 +166,66 @@ const getImgsFromUrls = (urls, target) => {
             'For now only inputs can be clicked to load the image on the active LoadImage node.',
           life: 5000,
         })
+      }
+    } else {
+      a.autoplay = true
+
+      a.muted = true
+      a.loop = true
+      a.onclick = (_e) => {
+        const selected = app.canvas.selected_nodes
+        if (selected && Object.keys(selected).length === 0) {
+          app.extensionManager.toast.add({
+            severity: 'warn',
+            summary: 'No node selected!',
+            detail:
+              "For now the only action when clicking videos in the sidebar is to set the video on all selected 'Load Video (Upload)' nodes.",
+            life: 5000,
+          })
+          return
+        }
+
+        for (const [_id, node] of Object.entries(app.canvas.selected_nodes)) {
+          updateImage(node, key)
+        }
+      }
     }
     imgs.push(a)
   }
   if (target !== undefined) {
-    target.append(...imgs)
+    if (options.prepend) target.prepend(...imgs)
+    else target.append(...imgs)
   }
   return imgs
 }
 
-const getUrls = async () => {
-  const count = await api.getSetting('mtb.io-sidebar.count')
+const getModes = async () => {
+  const inputs = await shared.runAction('getUserImageFolders')
+  return inputs
+}
+const getUrls = async (subfolder) => {
+  const count = (await api.getSetting('mtb.io-sidebar.count')) || 1000
   console.log('Sidebar count', count)
-  const inputs = await api.fetchApi('/mtb/actions', {
-    method: 'POST',
-    body: JSON.stringify({
-      name: 'getUserImages',
-      // mode, count, offset
-      args: [currentMode, count, offset, currentSort],
-    }),
-  })
-  const output = await inputs.json()
-  return output?.result || {}
+  if (currentMode === 'video') {
+    const output = await shared.runAction(
+      'getUserVideos',
+      256,
+      count,
+      offset,
+      currentSort,
+    )
+    return output || {}
+  }
+  const output = await shared.runAction(
+    'getUserImages',
+    currentMode,
+    count,
+    offset,
+    currentSort,
+    false,
+    subfolder,
+  )
+  return output || {}
 }
 
 //NOTE: do not load if using the old ui
@@ -187,21 +324,39 @@ if (window?.__COMFYUI_FRONTEND_VERSION__) {
             el.parentNode.style.overflowY = 'clip'
           }
 
-          const urls = await getUrls(currentMode)
+          const allModes = await getModes()
+          const input_modes = allModes.input.map((m) => `input - ${m}`)
+          const output_modes = allModes.output.map((m) => `output - ${m}`)
+          const urls = await getUrls()
           let imgs = {}
 
           const cont = makeElement('div.mtb_sidebar')
 
           const imgGrid = makeElement('div.mtb_img_grid')
-          const selector = makeSelect(['input', 'output'], currentMode)
+          const selector = makeSelect(
+            ['input', 'output', 'video', ...output_modes, ...input_modes],
+            currentMode,
+          )
 
           selector.addEventListener('change', async (e) => {
-            const newMode = e.target.value
-            const changed = newMode !== currentMode
+            let newMode = e.target.value
+            let changed = false
+            let newSub = ''
+            if (newMode !== 'input' && newMode !== 'output') {
+              if (newMode.startsWith('input - ')) {
+                newSub = newMode.replace('input - ', '')
+                newMode = 'input'
+              } else if (newMode.startsWith('output - ')) {
+                newSub = newMode.replace('output - ', '')
+                newMode = 'output'
+              }
+            }
+            changed = newMode !== currentMode || newSub !== subfolder
             currentMode = newMode
+            subfolder = newSub
             if (changed) {
               imgGrid.innerHTML = ''
-              const urls = await getUrls()
+              const urls = await getUrls(subfolder)
               if (urls) {
                 imgs = getImgsFromUrls(urls, imgGrid)
               }
@@ -220,7 +375,7 @@ if (window?.__COMFYUI_FRONTEND_VERSION__) {
             currentSort = newSort
             if (changed) {
               imgGrid.innerHTML = ''
-              const urls = await getUrls()
+              const urls = await getUrls(subfolder)
               if (urls) {
                 imgs = getImgsFromUrls(urls, imgGrid)
               }
@@ -229,7 +384,6 @@ if (window?.__COMFYUI_FRONTEND_VERSION__) {
 
           const sizeSlider = makeSlider(64, 1024, currentWidth, 1)
           imgTools.appendChild(orderSelect)
-
           imgTools.appendChild(sizeSlider)
 
           imgs = getImgsFromUrls(urls, imgGrid)
@@ -241,11 +395,16 @@ if (window?.__COMFYUI_FRONTEND_VERSION__) {
             }
           })
           handle = renderSidebar(el, cont, [selector, imgGrid, imgTools])
+          app.api.addEventListener('status', async () => {
+            if (currentMode !== 'output') return
+            updateOutputsGrid()
+          })
         },
         destroy: () => {
           if (handle) {
             handle.unregister()
             handle = undefined
+            app.api.removeEventListener('status')
           }
         },
       })

web/mtb_ui.js

@@ -184,6 +184,18 @@ ${inputs}
   )
 }
 
+/**
+ *  Wrap an element with a div
+ *
+ * @param {Object} [style] - CSS styles to apply to the element.
+ * @returns {HTMLElement} - The created DOM element.
+ */
+export const wrapElement = (element, style = {}) => {
+  const container = makeElement('div', style)
+  container.appendChild(element)
+  return container
+}
+
 /**
  * Creates a DOM element with optional styles, class, and id.
  *

web/mtb_widgets.js

@@ -21,7 +21,7 @@ import { infoLogger } from './comfy_shared.js'
 import { NumberInputWidget } from './numberInput.js'
 
 // NOTE: new widget types registered by MTB Widgets
-const newTypes = [/*'BOOL'*/ , 'COLOR', 'BBOX']
+const newTypes = [/*'BOOL'*/ 'COLOR', 'BBOX']
 
 const deprecated_nodes = {
   //  'Animation Builder':
@@ -536,21 +536,34 @@ export const MtbWidgets = {
             picker.type = 'color'
             picker.value = this.value
 
-            picker.style.position = 'absolute'
-            picker.style.left = '999999px' //(window.innerWidth / 2) + "px";
-            picker.style.top = '999999px' //(window.innerHeight / 2) + "px";
+            Object.assign(picker.style, {
+              position: 'fixed',
+              left: `${e.clientX}px`,
+              top: `${e.clientY}px`,
+              height: '0px',
+              width: '0px',
+              padding: '0px',
+              opacity: 0,
+            })
+
+            picker.addEventListener('blur', () => {
+              this.callback?.(this.value)
+              node.graph._version++
+              picker.remove()
+            })
+            picker.addEventListener('input', () => {
+              if (!picker.value) return
+
+              this.value = picker.value
+              app.canvas.setDirty(true)
+            })
 
             document.body.appendChild(picker)
 
-            picker.addEventListener('change', () => {
-              this.value = picker.value
-              this.callback?.(this.value)
-              node.graph._version++
-              node.setDirtyCanvas(true, true)
-              picker.remove()
+            requestAnimationFrame(() => {
+              picker.showPicker()
+              picker.focus()
             })
-
-            picker.click()
           }
         }
       }
@@ -659,8 +672,7 @@ const mtb_widgets = {
   init: async () => {
     infoLogger('Registering mtb.widgets')
     try {
-      const res = await api.fetchApi('/mtb/server-info')
-      const msg = await res.json()
+      const msg = await shared.getServerInfo()
       if (!window.MTB) {
         window.MTB = {}
       }
@@ -703,17 +715,11 @@ const mtb_widgets = {
           infoLogger('Enabled DEBUG mode')
         }
 
-        await api
-          .fetchApi('/mtb/server-info', {
-            method: 'POST',
-            body: JSON.stringify({
-              debug: value,
-            }),
-          })
-          .then((_response) => {})
-          .catch((error) => {
-            console.error('Error:', error)
-          })
+        try {
+          shared.setServerInfo({ debug: value })
+        } catch (err) {
+          console.error('Error:', err)
+        }
       },
     })
   },
@@ -1096,13 +1102,10 @@ const mtb_widgets = {
 
           const getStyle = async (node) => {
             try {
-              const getStyles = await api.fetchApi('/mtb/actions', {
-                method: 'POST',
-                body: JSON.stringify({
-                  name: 'getStyles',
-                  args: node.widgets?.[0].value ? node.widgets[0].value : '',
-                }),
-              })
+              const getStyles = await runAction(
+                'getStyles',
+                node.widgets?.[0].value ? node.widgets[0].value : '',
+              )
 
               const output = await getStyles.json()
               return output?.result
@@ -1202,6 +1205,8 @@ const mtb_widgets = {
         shared.setupDynamicConnections(nodeType, 'floats', 'FLOATS')
         break
       }
+      case 'Batch Sequence (mtb)':
+      case 'Batch Sequence Plus (mtb)':
       case 'Batch Merge (mtb)': {
         shared.setupDynamicConnections(nodeType, 'batches', 'IMAGE')
 
@@ -1278,23 +1283,141 @@ const mtb_widgets = {
         })
         break
       }
-      case 'Save Tensors (mtb)': {
+      case 'Scene Detect (mtb)': {
+        break
+      }
+      case 'Loop Start (mtb)': {
         const onDrawBackground = nodeType.prototype.onDrawBackground
-        nodeType.prototype.onDrawBackground = function (ctx, canvas) {
+        nodeType.prototype.onDrawBackground = function (...args) {
           const r = onDrawBackground
-            ? onDrawBackground.apply(this, arguments)
+            ? onDrawBackground.apply(this, args)
             : undefined
-          // // draw a circle on the top right of the node, with text inside
-          // ctx.fillStyle = "#fff";
-          // ctx.beginPath();
-          // ctx.arc(this.size[0] - this.node_width * 0.5, this.size[1] - this.node_height * 0.5, this.node_width * 0.5, 0, Math.PI * 2);
-          // ctx.fill();
+          const [ctx, /*canvas,*/ ..._rest] = args
+          if (this.flags.collapsed) return r
+          if (!this.computed_flow) {
+            const related = new Set([this.id])
+            const visited = new Set()
+            if (this.outputs[0].links) {
+              const initLink = this.outputs[0].links[0]
+              const { to: loopEnd } = shared.nodesFromLink(this, initLink)
+              const canReachEnd = (node, visited = new Set()) => {
+                if (node === loopEnd) return true
+                if (visited.has(node.id)) return false
+                visited.add(node.id)
+                for (const output of node.outputs || []) {
+                  if (!output.links) continue
+                  for (const linkId of output.links) {
+                    const { to: nextNode } = shared.nodesFromLink(node, linkId)
+                    if (!nextNode) continue
+                    if (canReachEnd(nextNode, visited)) {
+                      return true
+                    }
+                  }
+                }
+                return false
+              }
+              const traverseNodes = (node) => {
+                if (visited.has(node.id)) return
+                visited.add(node.id)
 
-          // ctx.fillStyle = "#000";
-          // ctx.textAlign = "center";
-          // ctx.font = "bold 12px Arial";
-          // ctx.fillText("Save Tensors", this.size[0] - this.node_width * 0.5, this.size[1] - this.node_height * 0.5);
+                // can reach the end
+                if (node !== this && node !== loopEnd && !canReachEnd(node)) {
+                  return
+                }
 
+                related.add(node.id)
+                for (const output of node.outputs || []) {
+                  if (!output.links) continue
+
+                  for (const linkId of output.links) {
+                    const { to: nextNode } = shared.nodesFromLink(node, linkId)
+                    if (!nextNode) continue
+
+                    traverseNodes(nextNode)
+                  }
+                }
+              }
+
+              traverseNodes(this)
+            }
+            this.related_to_flow = Array.from(related)
+            this.computed_flow = true
+          }
+          if (this.related_to_flow) {
+            ctx.save()
+            const points = []
+            const padding = 20
+
+            const graph = this.graph
+            const offset = this._pos
+
+            for (const nodeId of this.related_to_flow) {
+              const node = graph.getNodeById(nodeId)
+              if (!node) continue
+
+              const scale = 1.0
+              const x = node._pos[0] * scale - offset[0]
+              const y = node._pos[1] * scale - offset[1]
+              const width = node.size[0] * scale
+              const height = node.size[1] * scale
+              const scaledPadding = padding * scale
+              // console.log({ main: this, x, y, width, height })
+
+              points.push(
+                [x - scaledPadding, y - scaledPadding],
+                [x + width + scaledPadding, y - scaledPadding],
+                [x + width + scaledPadding, y + height + scaledPadding],
+                [x - scaledPadding, y + height + scaledPadding],
+              )
+            }
+            // console.log({ points })
+            const hull = shared.getConvexHull(points)
+
+            ctx.beginPath()
+
+            ctx.moveTo(hull[0][0], hull[0][1])
+            for (let i = 1; i < hull.length; i++) {
+              ctx.lineTo(hull[i][0], hull[i][1])
+            }
+
+            ctx.closePath()
+
+            ctx.fillStyle = 'rgba(255, 0, 0, 0.1)'
+            ctx.strokeStyle = 'rgba(255, 0, 0, 0.5)'
+            ctx.lineWidth = 2
+            ctx.fill()
+            ctx.stroke()
+
+            ctx.restore()
+          } else {
+            ctx.save()
+            ctx.fillStyle = 'red'
+            ctx.fillRect(-50, -50, this.size[0] + 100, this.size[1] + 100)
+            ctx.fillStyle = 'white'
+            ctx.font = 'bold 12px Arial'
+            ctx.fillText(
+              `pos: ${this.x}x${this.y}`,
+              this.size[0] / 2,
+              this.size[1],
+            )
+            ctx.fillText(
+              `size:${this._posSize}`,
+              this.size[0] / 2,
+              this.size[1] - 30,
+            )
+            ctx.fillText(
+              `dpi: ${window.devicePixelRatio}`,
+              this.size[0] / 2,
+              this.size[1] - 60,
+            )
+            ctx.fillText(
+              `next: ${graph.getNodeById(this.related_to_flow[1])._posSize}`,
+              this.size[0] / 2,
+              this.size[1] - 90,
+            )
+
+            ctx.restore()
+          }
           return r
         }
         break

web/note_plus.js

@@ -600,9 +600,9 @@ class NotePlus extends LiteGraph.LGraphNode {
     // this.updateHTML(this.html_widget.value) // widget is populated here since we called super
     // this.updateCSS(this.css_widget.value)
   }
-  onRemoved() {
-    infoLogger('Node removed', this.uuid)
-  }
+  // onRemoved() {
+  //   infoLogger('Node removed', this?.uuid)
+  // }
   getExtraMenuOptions() {
     const currentMode = this.edit_mode_widget.value
     const newMode = currentMode === 'html' ? 'markdown' : 'html'