39 Commits
Author SHA1 Message Date
kijai aa3f736239 Update nodes.py 2024-06-11 13:40:33 +03:00
Kijai 69181aa1cf Refactor v2 nodes 2024-05-29 14:49:01 +03:00
Kijai a222bafd8f Initial node for diffusers 0.28 marigold, including normal map gen 2024-05-28 17:12:22 +03:00
kijai beb2c3b5cb Add descriptions 2024-04-08 11:32:41 +03:00
kijai 80b3ee8948 smarter model loading 2024-04-08 11:13:29 +03:00
kijai bbb1550514 Add example workflows 2024-03-26 21:04:59 +02:00
kijai 9860e15978 LCM scheduler 2024-03-26 01:51:10 +02:00
kijai 1f0fd438cf add lcm model 2024-03-26 01:04:41 +02:00
Kijai d8fb90566c Add video processing node
Idea and original implementation from https://github.com/pablodawson/Marigold-Video
2024-03-14 13:55:36 +02:00
Kijai 42820b9c14 Add normalization option
Normalize to -1.0 - 1.0 by default, added boolean to test the difference (default on as it seems better)
2024-01-31 13:37:50 +02:00
kijai 983508c389 Attempt to support mps 2024-01-24 20:41:27 +02:00
kijai 91006653ea Add scheduler options 2023-12-27 15:08:22 +02:00
kijai be0db21630 Update nodes.py 2023-12-22 20:30:31 +02:00
kijai 55fb000899 Merge branch 'main' of https://github.com/kijai/ComfyUI-Marigold 2023-12-22 19:56:06 +02:00
kijai 7817260198 Make model autodownload if not found (safetensors only) 2023-12-22 19:56:03 +02:00
Kijai de927867a9 Fix output dtype 2023-12-19 11:06:29 +02:00
kijai 17f8db90a1 Reload model if dtype changes 2023-12-17 14:37:59 +02:00
kijai b7bf6f745f fix RemapDepth on older pytorch 2023-12-15 22:30:41 +02:00
kijai 6b143fbee2 fix cv2 exr export 2023-12-15 12:15:35 +02:00
kijai b5de3e61a8 Attempt to support cv2 for EXR 2023-12-15 11:15:27 +02:00
kijai 74cded2e18 Keep the model from reloading.. for real this time 2023-12-14 23:02:43 +02:00
kijai 60345a21c7 make fp16 optional
...but it takes twice the VRAM to use without
2023-12-14 22:24:29 +02:00
kijai b33af683ef Attempt to fix empty embed loading path 2023-12-14 21:39:27 +02:00
kijai 2f63519c4d Add RemapDepth node 2023-12-14 17:59:04 +02:00
Fannovel16 1a5017007a Cache empty text embedding (#7)
* Cache empty text embedding to reduce a bit of time and avoid CLIP loading

* It is only 9MB so moving to global won't affect RAM much

* Complete
2023-12-14 17:31:13 +02:00
kijai f5bc8dea80 Update nodes.py 2023-12-14 03:01:23 +02:00
kijai b8ecf8a29d Add openEXR save node 2023-12-14 01:27:26 +02:00
kijai ce1ba31fa1 Update nodes.py 2023-12-13 22:27:31 +02:00
kijai d5529f69d5 remove prints 2023-12-13 22:18:58 +02:00
kijai 30049bc659 Separate colormap to it's own node 2023-12-13 22:13:02 +02:00
kijai 77fad21907 Add colorize option 2023-12-13 20:57:19 +02:00
kijai 6a49443b2d Add batch processing to the n_repeats 2023-12-13 19:37:42 +02:00
kijai f01ecfd078 Add option to unload the model after generation 2023-12-13 15:25:39 +02:00
kijai 59fd6f62ce Add diffusers folder as model location option 2023-12-13 09:44:38 +02:00
kijai 315c8a5b95 Add progress bar 2023-12-13 00:57:55 +02:00
kijai c27a1d56be run in fp16 2023-12-12 23:59:42 +02:00
kijai a3d3f849b3 add requirements 2023-12-12 22:40:20 +02:00
kijai 6ce5230978 update 2023-12-12 20:07:32 +02:00
kijai e6e3604144 Initial commit 2023-12-12 18:48:21 +02:00