Version 3.4.0

* Add Anima region attention support

* Added adaptation acknowledgement for Anima Attention couple implementation

* Address Anima attention cleanup feedback

.github/workflows/publish.yml

@@ -11,11 +11,12 @@ jobs:
   publish-node:
     name: Publish Custom Node to registry
     runs-on: ubuntu-latest
+    if: ${{ github.repository_owner == 'Acly' }}
     steps:
       - name: Check out code
         uses: actions/checkout@v4
       - name: Publish Custom Node
-        uses: Comfy-Org/publish-node-action@main
+        uses: Comfy-Org/publish-node-action@v1
         with:
           ## Add your own personal access token to your Github Repository secrets and reference it here.
           personal_access_token: ${{ secrets.REGISTRY_ACCESS_TOKEN }}

README.md

@@ -13,9 +13,9 @@ Provides nodes and API geared towards using ComfyUI as a backend for external to
 ## <a id="images" href="#toc">Sending and receiving images</a>
 
 ComfyUI exchanges images via the filesystem. This requires a
-multi-step process (upload images, prompt, download images), is rather
+multi-step process (upload images, prompt, download images), which
-inefficient, and invites a whole class of potential issues. It's also unclear
+invites a whole class of potential issues you might not want to deal with.
-at which point those images will get cleaned up if ComfyUI is used
+It's also unclear at which point those images will get cleaned up if ComfyUI is used
 via external tools.
 
 ### Load Image (Base64)
@@ -33,7 +33,7 @@ Loads a mask (single channel) from a PNG embedded into the prompt as base64 stri
 ### Send Image (WebSocket)
 
 Sends an output image over the client WebSocket connection as PNG binary data.
-* Inputs: the image (RGB or RGBA)
+* Inputs: the image (RGB or RGBA), supports batches
 
 This will first send one binary message for each image in the batch via WebSocket:
 ```
@@ -44,6 +44,47 @@ That is two 32-bit integers (big endian) with values 1 and 2 followed by the PNG
 {'type': 'executed', 'data': {'node': '<node ID>', 'output': {'images': [{'source': 'websocket', 'content-type': 'image/png', 'type': 'output'}, ...]}, 'prompt_id': '<prompt ID>}}
 ```
 
+### Load Image from Cache
+
+Loads an image or mask that has been uploaded previously into the workflow.
+Uploaded images are temporarily stored in RAM rather than written to disk. This
+method has less overhead compared to embedding images as base64 into the prompt,
+but is more complex to implement.
+* Inputs: id of an image that was uploaded previously
+* Outputs: image (RGB) and mask (A of RGBA input, or first channel if no alpha present).
+
+To upload an image, upload the _bytes_ of a PNG via a HTTP PUT request to
+`/api/etn/image/{id}`. JPEG or other formats also work. Choose any `id` which
+does not clash with other images you upload, and reference it in the node. The
+request returns `201` if the image was uploaded and `200` if it was already
+cached.
+
+### Save Image to Cache
+
+Stores an output image in RAM temporarily and allows retrieval over HTTP.
+This is typically faster than WebSocket, especially for large images.
+* Inputs: the image (RGB or RGBA). Batches are supported.
+
+This node will send a JSON message over WebSocket when an image is ready:
+```json
+{
+  "type": "executed",
+  "data": {
+    "node": "<node ID>",
+    "output": {
+      "images": [
+        {"source": "http", "id": "<image ID>", "content-type": "image/png", "type": "output"}
+      ]
+    },
+    "prompt_id": "prompt ID"
+  }
+}
+```
+
+To download the images, send a HTTP GET request to `/api/etn/image/{id}` with
+the image IDs from the message. Images will be cached for a few minutes.
+
+
 ## <a id="regions" href="#toc">Regions</a>
 
 These nodes implement attention masking for arbitrary number of image regions. Text prompts only apply to the masked area.
@@ -163,7 +204,9 @@ There are various types of models that can be loaded as checkpoint, LoRA, Contro
 
 #### Paramters
 * `folder_name`: sub-directory in ComfyUI's models folder.
-  Supported model types: `checkpoints`, `diffusion_models`
+  Supported model types: `checkpoints`, `diffusion_models`, `unet`, `unet_gguf`
+* `limit=n`: (query parameter, optional) inspect at `n` models
+* `offset=i`: (query parameter, optional) start with the `i`th model
 
 #### Output
 Lists available models with additional classification info:
@@ -172,14 +215,38 @@ Lists available models with additional classification info:
     "checkpoint_file.safetensors": {
         "base_model": "sd15",
         "is_inpaint": false,
-        "is_refiner": false
+        "type": "eps"
     },
     ...
 }
 ```
-Possible values for base model: `sd15, sd20, sd21, sd3, sdxl, ssd1b, svd, cascade-b, cascade-c, aura-flow, hunyuan-dit, flux, flux-schnell`
+Possible values for base model: `sd15, sd20, sd21, sd3, sdxl, sdxl-refiner, ssd1b, svd, cascade-b, cascade-c, aura-flow, hunyuan-dit, flux, flux-schnell, flux2, lumina2, z-image, chroma, qwen-image`
+
+If base model is `sdxl`, the `type` attribute is set with possible values: `eps, edm, v-prediction, v-prediction-edm`
+
+Detection supports quantized models:
+* GGUF: if the `gguf` module is installed, .gguf files are detected and will set the `quant` field
+* Nunchaku: SVDQuant models are detected and will set the `quant` field to `svdq`
+
+Returns an entry `{"base_model": "unknown"}` for models with unknown format or which do not match any of the known base models.
+
+
+#### Pagination
+The query parameters limit and offset allow inspecting a subset of models per request.
+Usually inspection is quite fast (it only looks at model headers), but it can be slow
+in some cases due to anti-virus or slow harddrives.
+```
+GET /api/etn/model_info/checkpoints?limit=10&offset=20
+```
+This will return at most 10 models, starting with the 20th model in the list.
+It also returns a special `_meta` entry in the output JSON:
+```json
+{
+    "checkpoint_20.safetensors": { ... },
+    "_meta": { "offset": 20, "count": 1, "total": 21 }
+}
+```
 
-The entry is `{"base_model": "unknown"}` for models which are not in safetensors format or do not match any of the known base models.
 
 ### GET /api/etn/languages
 
@@ -226,3 +293,9 @@ git clone https://github.com/Acly/comfyui-tooling-nodes.git
 ```
 
 Restart ComfyUI and the nodes are functional.
+
+
+## Acknowledgements
+
+* Region nodes adapted from [laksjdjf/cgem156-ComfyUI](https://github.com/laksjdjf/cgem156-ComfyUI/blob/main/scripts/attention_couple/node.py)
+* Control nodes adapted from [kohya-ss/ComfyUI-Anima-LLLite](https://github.com/kohya-ss/ComfyUI-Anima-LLLite)

__init__.py

@@ -1,53 +1,57 @@
-from . import api, nodes, tile, region, nsfw, translation, krita
+from comfy_api.latest import ComfyExtension, io
+
+from . import api as api
+from . import control, krita, nodes, region, tile, translation
+
+
+class ExternalToolingNodes(ComfyExtension):
+    async def get_node_list(self) -> list[type[io.ComfyNode]]:
+        node_list = [
+            nodes.LoadImageCache,
+            nodes.SaveImageCache,
+            nodes.LoadImageBase64,
+            nodes.LoadMaskBase64,
+            nodes.SendImageWebSocket,
+            nodes.ApplyMaskToImage,
+            nodes.ReferenceImage,
+            nodes.ApplyReferenceImages,
+            tile.CreateTileLayout,
+            tile.ExtractImageTile,
+            tile.ExtractMaskTile,
+            tile.GenerateTileMask,
+            tile.MergeImageTile,
+            region.BackgroundRegion,
+            region.DefineRegion,
+            region.ListRegionMasks,
+            region.AttentionMask,
+            translation.Translate,
+            krita.KritaOutput,
+            krita.KritaSendText,
+            krita.KritaCanvas,
+            krita.KritaSelection,
+            krita.KritaImageLayer,
+            krita.KritaMaskLayer,
+            krita.Parameter,
+            krita.KritaStyle,
+            krita.KritaStyleAndPrompt,
+            control.ControlApply,
+            control.ControlLoad,
+        ]
+        try:  # see #66
+            from . import nsfw
+
+            node_list.append(nsfw.NSFWFilter)
+        except (ImportError, ModuleNotFoundError):
+            import traceback
+
+            print("[comfyui-tooling-nodes] WARNING: Could not import all nodes.")
+            traceback.print_exc()
+
+        return node_list
+
+
+async def comfy_entrypoint():
+    return ExternalToolingNodes()
+
 
-NODE_CLASS_MAPPINGS = {
-    "ETN_LoadImageBase64": nodes.LoadImageBase64,
-    "ETN_LoadMaskBase64": nodes.LoadMaskBase64,
-    "ETN_SendImageWebSocket": nodes.SendImageWebSocket,
-    "ETN_CropImage": nodes.CropImage,
-    "ETN_ApplyMaskToImage": nodes.ApplyMaskToImage,
-    "ETN_TileLayout": tile.TileLayout,
-    "ETN_ExtractImageTile": tile.ExtractImageTile,
-    "ETN_ExtractMaskTile": tile.ExtractMaskTile,
-    "ETN_GenerateTileMask": tile.GenerateTileMask,
-    "ETN_MergeImageTile": tile.MergeImageTile,
-    "ETN_BackgroundRegion": region.BackgroundRegion,
-    "ETN_DefineRegion": region.DefineRegion,
-    "ETN_ListRegionMasks": region.ListRegionMasks,
-    "ETN_AttentionMask": region.AttentionMask,
-    "ETN_NSFWFilter": nsfw.NSFWFilter,
-    "ETN_Translate": translation.Translate,
-    "ETN_KritaOutput": krita.KritaOutput,
-    "ETN_KritaCanvas": krita.KritaCanvas,
-    "ETN_KritaSelection": krita.KritaSelection,
-    "ETN_KritaImageLayer": krita.KritaImageLayer,
-    "ETN_KritaMaskLayer": krita.KritaMaskLayer,
-    "ETN_Parameter": krita.Parameter,
-    "ETN_KritaStyle": krita.KritaStyle,
-}
-NODE_DISPLAY_NAME_MAPPINGS = {
-    "ETN_LoadImageBase64": "Load Image (Base64)",
-    "ETN_LoadMaskBase64": "Load Mask (Base64)",
-    "ETN_SendImageWebSocket": "Send Image (WebSocket)",
-    "ETN_CropImage": "Crop Image",
-    "ETN_ApplyMaskToImage": "Apply Mask to Image",
-    "ETN_TileLayout": "Create Tile Layout",
-    "ETN_ExtractImageTile": "Extract Image Tile",
-    "ETN_ExtractMaskTile": "Extract Mask Tile",
-    "ETN_MergeImageTile": "Merge Image Tile",
-    "ETN_GenerateTileMask": "Generate Tile Mask",
-    "ETN_BackgroundRegion": "Background Region",
-    "ETN_DefineRegion": "Define Region",
-    "ETN_ListRegionMasks": "List Region Masks",
-    "ETN_AttentionMask": "Regions Attention Mask",
-    "ETN_NSFWFilter": "NSFW Filter",
-    "ETN_Translate": "Translate Text",
-    "ETN_KritaOutput": "Krita Output",
-    "ETN_KritaCanvas": "Krita Canvas",
-    "ETN_KritaSelection": "Krita Selection",
-    "ETN_KritaImageLayer": "Krita Image Layer",
-    "ETN_KritaMaskLayer": "Krita Mask Layer",
-    "ETN_Parameter": "Parameter",
-    "ETN_KritaStyle": "Krita Style",
-}
 WEB_DIRECTORY = "./js"

api.py

@@ -1,19 +1,21 @@
 from __future__ import annotations
 from aiohttp import web
-from typing import NamedTuple
+from typing import Any, NamedTuple
 from pathlib import Path
 import json
 import traceback
 import re
 import logging
+import itertools
 
-from comfy import model_detection, supported_models
+from comfy import model_detection
 import comfy.utils
 import folder_paths
 import server
 
 from .translation import available_languages, translate
 from .krita import WorkflowExchange
+from .nodes import image_cache
 
 input_block_name = "model.diffusion_model.input_blocks.0.0.weight"
 
@@ -22,7 +24,7 @@ model_names = {
     "SD20": "sd20",
     "SD21UnclipL": "sd21",
     "SD21UnclipH": "sd21",
-    "SDXLRefiner": "sdxl",
+    "SDXLRefiner": "sdxl-refiner",
     "SDXL": "sdxl",
     "SSD1B": "ssd1b",
     "SVD_img2vid": "svd",
@@ -33,7 +35,37 @@ model_names = {
     "HunyuanDiT": "hunyuan-dit",
     "HunyuanDiT1": "hunyuan-dit",
     "Flux": "flux",
+    "FluxInpaint": "flux",
     "FluxSchnell": "flux-schnell",
+    "GenmoMochi": "mochi",
+    "LTXV": "ltxv",
+    "HunyuanVideo": "hunyuan-video",
+    "CosmosT2V": "cosmos",
+    "CosmosI2V": "cosmos",
+    "CosmosT2IPredict2": "cosmos-predict2",
+    "CosmosI2VPredict2": "cosmos-predict2",
+    "ZImage": "z-image",
+    "Lumina2": "lumina2",
+    "WAN21_T2V": "wan21",
+    "WAN21_I2V": "wan21",
+    "WAN21_FunControl2V": "wan21-fun",
+    "WAN21_Vace": "wan21-vace",
+    "WAN21_Camera": "wan21-camera",
+    "HiDream": "hi-dream",
+    "Chroma": "chroma",
+    "ACEStep": "ace-step",
+    "Omnigen2": "omnigen2",
+    "QwenImage": "qwen-image",
+    "QwenImage21": "qwen-image21",
+    "ErnieImage": "ernie-image",
+    "Flux2": "flux2",
+    "Anima": "anima",
+    "Krea2": "krea2",
+}
+
+gguf_architectures = {
+    "sd1": "sd15",
+    "qwen_image": "qwen-image",
 }
 
 
@@ -48,7 +80,7 @@ class FakeTensor(NamedTuple):
             return d
 
 
-def inspect_diffusion_model(filename: str, prefix: str | None, model_type: str):
+def inspect_safetensors(filename: str, model_type: str, is_checkpoint: bool):
     try:
         # Read header of safetensors file
         path = folder_paths.get_full_path(model_type, filename)
@@ -62,8 +94,10 @@ def inspect_diffusion_model(filename: str, prefix: str | None, model_type: str):
                     cfg[key] = FakeTensor.from_dict(cfg[key])
 
             # Reuse Comfy's model detection
-            if prefix is None:
+            prefix = model_detection.unet_prefix_from_state_dict(cfg)
-                prefix = model_detection.unet_prefix_from_state_dict(cfg)
+            if not is_checkpoint:
+                cfg = comfy.utils.state_dict_prefix_replace(cfg, {prefix: ""}, filter_keys=False)
+                prefix = ""
             try:  # latest ComfyUI takes 2 args
                 unet_config = model_detection.detect_unet_config(cfg, prefix)
             except TypeError as e:  # older ComfyUI versions take 3 args
@@ -76,30 +110,161 @@ def inspect_diffusion_model(filename: str, prefix: str | None, model_type: str):
                 input_count = 4
 
             # Find a matching base model depending on unet config
-            base_model = model_detection.model_config_from_unet_config(unet_config)
+            base_model = None
-            if base_model is None:
+            model_type = None
+            model_quant = None
+
+            # Check if it's a Nunchaku SVDQ model by inspecting metadata
+            raw_name = detect_svdq(cfg)
+            if raw_name:
+                model_quant = "svdq"
+            # Otherwise try ComfyUI's model detection
+            elif unet_config is not None:
+                base_model = model_detection.model_config_from_unet_config(unet_config)
+                if base_model:
+                    raw_name = base_model.__class__.__name__
+                    if raw_name == "SDXL":
+                        model_type = base_model.model_type(cfg).name.lower().replace("_", "-")
+                    if raw_name == "Flux2":
+                        hidden_size = unet_config.get("hidden_size", 0)
+                        model_type = {3072: "klein-4b", 4096: "klein-9b"}.get(hidden_size, "dev")
+
+            if not raw_name:
                 return {"base_model": "unknown"}
 
-            base_model_class = base_model.__class__
+            base_model_name = model_names.get(raw_name, "unknown")
-            base_model_name = model_names.get(base_model_class.__name__, "unknown")
+            result: dict[str, Any] = {"base_model": base_model_name}
-            return {
+            result["is_inpaint"] = (
-                "base_model": base_model_name,
+                base_model_name in ["sd15", "sdxl"] and input_count > 4
-                "is_inpaint": base_model_name in ["sd15", "sdxl"] and input_count > 4,
+            ) or raw_name == "FluxInpaint"
-                "is_refiner": base_model_class is supported_models.SDXLRefiner,
+            if model_quant:
-            }
+                result["quant"] = model_quant
+            if model_type:
+                result["type"] = model_type
+            elif "T2I" in raw_name:
+                result["type"] = "t2i"
+            elif "I2V" in raw_name:
+                result["type"] = "i2v"
+            elif "T2V" in raw_name:
+                result["type"] = "t2v"
+            elif "Control2V" in raw_name:
+                result["type"] = "control2v"
+            return result
         return {"base_model": "unknown"}
+    except Exception as e:
+        print("[comfyui-tooling-nodes] Error inspecting file", filename)
+        traceback.print_exc()
+        return {"base_model": "unknown", "error": f"Failed to detect base model: {e}"}
+
+
+def detect_svdq(cfg: dict) -> str | None:
+    if md := cfg.get("__metadata__"):
+        if comfy_config := md.get("comfy_config"):
+            if isinstance(comfy_config, str):
+                comfy_config = json.loads(comfy_config)
+            if model_class := comfy_config.get("model_class"):
+                return model_class
+
+        match md.get("model_class"):
+            case "NunchakuFluxTransformer2dModel":
+                return "Flux"
+            case "NunchakuQwenImageTransformer2DModel":
+                return "QwenImage"
+            case "NunchakuZImageTransformer2DModel":
+                return "ZImage"
+    return None
+
+
+def inspect_gguf(filename: str, model_type: str):
+    try:
+        import gguf
+    except ImportError:
+        return {"base_model": "unknown", "error": "GGUF module not found"}
+
+    try:
+        path = folder_paths.get_full_path(model_type, filename)
+        if path is None:
+            raise Exception(f"Could not find full path for {model_type}/{filename}")
+
+        reader = gguf.GGUFReader(path)
+        arch_field = reader.get_field("general.architecture")
+        if arch_field is not None:
+            if len(arch_field.types) != 1 or arch_field.types[0] != gguf.GGUFValueType.STRING:
+                raise TypeError(
+                    f"Bad type for GGUF general.architecture key: expected string, got {arch_field.types!r}"
+                )
+            arch_str = str(arch_field.parts[arch_field.data[-1]], encoding="utf-8")
+        else:  # stable-diffusion.cpp, requires conversion. not handled for now
+            return {"base_model": "flux", "is_inpaint": False}
+
+        if arch_str == "flux" and any(
+            t.name.startswith("distilled_guidance_layer")
+            for t in itertools.islice(reader.tensors, 5)
+        ):
+            arch_str = "chroma"
+
+        # Detect Z-Image (modified Lumina2)
+        if arch_str == "lumina2":
+            for t in reader.tensors:
+                if t.name == "cap_embedder.1.bias" and t.shape[0] == 3840:
+                    arch_str = "z-image"
+                    break
+
+        # Detect Flux variants
+        result_type = None
+        if arch_str == "flux":
+            for t in reader.tensors:
+                if t.name.startswith("distilled_guidance_layer"):
+                    arch_str = "chroma"
+                    break
+                elif t.name == "double_stream_modulation_img.lin.weight":
+                    arch_str = "flux2"
+                    if t.shape[0] == 3072:
+                        result_type = "klein-4b"
+                    elif t.shape[0] == 4096:
+                        result_type = "klein-9b"
+                    break
+
+        result = {
+            "base_model": gguf_architectures.get(arch_str, arch_str),
+            "is_inpaint": False,
+        }
+        if result_type is not None:
+            result["type"] = result_type
+        try:
+            if file_type := reader.get_field("general.file_type"):
+                result["quant"] = file_type.contents().lower()
+        except Exception:
+            result["quant"] = "gguf"
+        return result
+
     except Exception as e:
         # traceback.print_exc()
         return {"base_model": "unknown", "error": f"Failed to detect base model: {e}"}
 
 
-def inspect_models(model_type: str):
+def inspect_diffusion_model(filename: str, model_type: str, is_checkpoint: bool):
+    if filename.endswith(".gguf"):
+        return inspect_gguf(filename, model_type)
+    return inspect_safetensors(filename, model_type, is_checkpoint)
+
+
+def inspect_models(model_type: str, params: dict[str, str]):
     try:
-        prefix = "" if model_type in ("unet", "diffusion_models") else None
+        try:
+            files = folder_paths.get_filename_list(model_type)
+        except KeyError:
+            return web.json_response({"error": f"Model folder not found: {model_type}"})
+        limit = int(params.get("limit", "1000"))
+        offset = int(params.get("offset", "0"))
+        files_range = files[offset : offset + limit]
+        is_checkpoint = model_type == "checkpoints"
         info = {
-            filename: inspect_diffusion_model(filename, prefix, model_type)
+            filename: inspect_diffusion_model(filename, model_type, is_checkpoint)
-            for filename in folder_paths.get_filename_list(model_type)
+            for filename in files_range
         }
+        if "limit" in params:
+            info["_meta"] = dict(offset=offset, count=len(files_range), total=len(files))
         return web.json_response(info)
     except Exception as e:
         traceback.print_exc()
@@ -128,25 +293,28 @@ def has_invalid_filename(filename: str):
     return None
 
 
+async def image_sender(data: bytes):
+    mem = memoryview(data)
+    csize = 2**14
+    for i in range(0, len(mem), csize):
+        yield mem[i : i + csize]
+
+
 _server: server.PromptServer | None = getattr(server.PromptServer, "instance", None)
 if _server is not None:
-
     _workflow_exchange = WorkflowExchange(_server)
 
     @_server.routes.get("/api/etn/model_info/{folder_name}")
     async def model_info(request: web.Request):
         folder_name = request.match_info.get("folder_name", "checkpoints")
-        if error := has_invalid_folder_name(folder_name):
+        error = has_invalid_folder_name(folder_name)
+        if error is not None:
             return error
-        return inspect_models(folder_name)
+        return inspect_models(folder_name, request.rel_url.query)
 
     @_server.routes.get("/api/etn/model_info")
     async def api_model_info(request):
-        return inspect_models("checkpoints")
+        return inspect_models("checkpoints", request.rel_url.query)
-
-    @_server.routes.get("/etn/model_info")
-    async def api_model_info(request):
-        return inspect_models("checkpoints")
 
     @_server.routes.get("/api/etn/languages")
     async def languages(request):
@@ -166,14 +334,60 @@ if _server is not None:
         except Exception as e:
             return web.json_response(dict(error=str(e)), status=500)
 
+    @_server.routes.get("/api/etn/image/{id}")
+    async def get_image(request: web.Request):
+        try:
+            id = request.match_info.get("id", "")
+            data, content_type = image_cache.get(id)
+            if data is None or content_type is None:
+                return web.json_response(dict(error="Image not found"), status=404)
+            response = web.Response(
+                body=image_sender(data),
+                content_type=content_type,
+                headers={"Content-Length": str(len(data))},
+            )
+            return response
+        except Exception as e:
+            return web.json_response(dict(error=str(e)), status=500)
+
+    async def put_image(request: web.Request):
+        try:
+            id = request.match_info.get("id", "")
+            if id in image_cache:
+                await request.release()  # Consume and discard the data to avoid connection abort
+                return web.json_response(dict(status="cached"), status=200)
+
+            content_type = request.headers.get("Content-Type", "application/octet-stream")
+            data = bytearray()
+            async for chunk, _ in request.content.iter_chunks():
+                data.extend(chunk)
+
+            image_cache.insert(id, bytes(data), content_type)
+            return web.json_response(dict(status="success"), status=201)
+        except Exception as e:
+            return web.json_response(dict(error=str(e)), status=500)
+
+    async def _put_image_expect_handler(request: web.Request):
+        if request.match_info.get("id", "") in image_cache:
+            # Skip "100 Continue" since we don't need the data, return 200 immediately.
+            return web.json_response(dict(status="cached"), status=200)
+        # otherwise run default aiohttp handler
+        return None
+
+    _server.app.router.add_route(
+        "PUT", "/api/etn/image/{id}", put_image, expect_handler=_put_image_expect_handler
+    )
+
     @_server.routes.put("/api/etn/upload/{folder_name}/{filename}")
     async def upload(request: web.Request):
         folder_name = request.match_info.get("folder_name", "")
-        if error := has_invalid_folder_name(folder_name):
+        error = has_invalid_folder_name(folder_name)
+        if error is not None:
             return error
 
         filename = request.match_info.get("filename", "")
-        if error := has_invalid_filename(filename):
+        error = has_invalid_filename(filename)
+        if error is not None:
             return error
 
         try:
@@ -182,7 +396,9 @@ if _server is not None:
 
             folder = Path(folder_paths.folder_names_and_paths[folder_name][0][0])
             total_size = int(request.headers.get("Content-Length", "0"))
-            logging.info(f"Uploading {filename} ({total_size/(1024**2):.1f} MB) to {folder} folder")
+            logging.info(
+                f"Uploading {filename} ({total_size / (1024**2):.1f} MB) to {folder} folder"
+            )
 
             with open(folder / filename, "wb") as f:
                 async for chunk, _ in request.content.iter_chunks():

control.py

@@ -0,0 +1,859 @@
+"""ControlNet-LLLite for Anima (DiT) — ComfyUI port (v2 architecture).
+
+Adapted from kohya-ss/ComfyUI-Anima-LLLite
+https://github.com/kohya-ss/ComfyUI-Anima-LLLite
+Apache-2.0 license
+
+Adapted from kohya-ss/sd-scripts. The on-disk weight format is the v2
+named-key format (per-module key prefix = lllite_name, shared encoder under
+``lllite_conditioning1.*``, depth embedding split per-module as
+``{name}.depth_embed``); legacy ``lllite_modules.*`` files are rejected.
+
+Differences vs. the sd-scripts reference (``networks/control_net_lllite_anima.py``):
+  * No dependency on ``library.utils`` — uses stdlib logging.
+  * Module discovery filters the LLM-Adapter sub-tree by class identity in
+    addition to the path-based check (ComfyUI ships two distinct ``Attention``
+    classes that share the bare class name).
+  * ``LLLiteModuleDiT`` keeps a ``restore()`` method (and an idempotent
+    ``apply_to()``); ComfyUI patches/unpatches the original Linear around
+    every sampler call via ``set_model_unet_function_wrapper``.
+  * Forward pass casts ``x`` and ``cond_emb`` to the LLLite parameter dtype
+    so autocast / mixed-precision flows that hand us a different dtype than
+    the LLLite weights still work.
+  * CFG batch-size and sequence-length mismatches fall back to identity
+    instead of asserting, so a slightly-off cond image cannot abort sampling.
+  * The training-side ``AnimaControlNetLLLiteWrapper`` is omitted; ComfyUI
+    integrates via ``model_function_wrapper`` in nodes.py instead.
+"""
+
+from __future__ import annotations
+
+from copy import copy
+import logging
+import os
+from dataclasses import dataclass
+from typing import Any
+
+import folder_paths
+import safetensors
+import safetensors.torch
+import torch
+import torch.nn.functional as F
+from comfy.model_patcher import ModelPatcher
+from comfy_api.latest import io
+from torch import nn
+
+logger = logging.getLogger("comfyui-tooling-nodes")
+
+
+# Class names of the modules that LLLite injects into. The LLM-Adapter uses
+# a different ``Attention`` class with the same bare name; we filter it by
+# path (``llm_adapter`` in the qualified name) and by the ``is_selfattn``
+# attribute presence.
+TARGET_ATTENTION_CLASS = "Attention"
+TARGET_MLP_CLASS = "GPT2FeedForward"
+LLM_ADAPTER_NAME = "llm_adapter"
+
+LLLITE_ARCH_VERSION = "2"
+
+
+# ----------------------------------------------------------------------------
+# target_layers: atomic specifiers and presets
+# ----------------------------------------------------------------------------
+
+ATOMIC_SPECIFIERS: tuple[str, ...] = (
+    "self_attn_q_pre",
+    "self_attn_kv_pre",
+    "cross_attn_q_pre",
+    "mlp_fc1_pre",
+)
+
+PRESETS: dict = {
+    "self_attn_q": ("self_attn_q_pre",),
+    "self_attn_qkv": ("self_attn_q_pre", "self_attn_kv_pre"),
+    "self_attn_qkv_cross_q": ("self_attn_q_pre", "self_attn_kv_pre", "cross_attn_q_pre"),
+}
+
+
+def parse_target_layers(spec: str) -> tuple[str, ...]:
+    """Resolve a ``target_layers`` spec to a canonical atomic tuple.
+
+    Accepts a preset name (``"self_attn_qkv"``) or a comma-separated list of
+    atomic specifiers (``"self_attn_q_pre,mlp_fc1_pre"``). Returns the atomics
+    in ``ATOMIC_SPECIFIERS`` order with duplicates removed.
+    """
+    if not isinstance(spec, str):
+        raise TypeError(f"target_layers must be str, got {type(spec).__name__}")
+    spec = spec.strip()
+    if not spec:
+        raise ValueError("target_layers spec is empty")
+
+    if spec in PRESETS:
+        parts = list(PRESETS[spec])
+    else:
+        parts = [p.strip() for p in spec.split(",") if p.strip()]
+        bad = [p for p in parts if p not in ATOMIC_SPECIFIERS]
+        if bad:
+            raise ValueError(
+                f"unknown target_layers atomic specifier(s): {bad}. "
+                f"valid atomic={list(ATOMIC_SPECIFIERS)}, presets={list(PRESETS)}"
+            )
+
+    return tuple(a for a in ATOMIC_SPECIFIERS if a in parts)
+
+
+# ----------------------------------------------------------------------------
+# Conditioning1 trunk (v2)
+# ----------------------------------------------------------------------------
+
+
+def _gn(channels: int) -> nn.GroupNorm:
+    g = 8
+    while g > 1 and channels % g != 0:
+        g //= 2
+    return nn.GroupNorm(g, channels)
+
+
+class _ResBlock(nn.Module):
+    def __init__(self, ch: int):
+        super().__init__()
+        self.norm1 = _gn(ch)
+        self.conv1 = nn.Conv2d(ch, ch, kernel_size=3, padding=1)
+        self.norm2 = _gn(ch)
+        self.conv2 = nn.Conv2d(ch, ch, kernel_size=3, padding=1)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        h = self.conv1(F.silu(self.norm1(x)))
+        h = self.conv2(F.silu(self.norm2(h)))
+        return x + h
+
+
+ASPP_DEFAULT_DILATIONS: tuple[int, ...] = (1, 2, 4, 8)
+
+
+class _ASPP(nn.Module):
+    def __init__(self, ch: int, dilations: tuple[int, ...] = ASPP_DEFAULT_DILATIONS):
+        super().__init__()
+        assert len(dilations) >= 1, "ASPP needs at least one dilation"
+        branches = []
+        for d in dilations:
+            if d == 1:
+                conv = nn.Conv2d(ch, ch, kernel_size=1)
+            else:
+                conv = nn.Conv2d(ch, ch, kernel_size=3, padding=d, dilation=d)
+            branches.append(nn.Sequential(conv, _gn(ch), nn.SiLU()))
+        self.branches = nn.ModuleList(branches)
+
+        self.global_pool = nn.AdaptiveAvgPool2d(1)
+        self.global_conv = nn.Sequential(nn.Conv2d(ch, ch, kernel_size=1), _gn(ch), nn.SiLU())
+
+        n_branches = len(dilations) + 1
+        self.proj = nn.Sequential(nn.Conv2d(ch * n_branches, ch, kernel_size=1), _gn(ch), nn.SiLU())
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        h, w = x.shape[-2:]
+        outs = [b(x) for b in self.branches]
+        g = self.global_conv(self.global_pool(x))
+        g = F.interpolate(g, size=(h, w), mode="bilinear", align_corners=False)
+        outs.append(g)
+        return self.proj(torch.cat(outs, dim=1))
+
+
+class _Conditioning1(nn.Module):
+    def __init__(
+        self,
+        cond_dim: int,
+        cond_emb_dim: int,
+        n_resblocks: int,
+        use_aspp: bool = False,
+        aspp_dilations: tuple[int, ...] = ASPP_DEFAULT_DILATIONS,
+        cond_in_channels: int = 3,
+    ):
+        super().__init__()
+        assert cond_dim % 2 == 0, f"cond_dim must be even, got {cond_dim}"
+        assert cond_in_channels >= 1, f"cond_in_channels must be >= 1, got {cond_in_channels}"
+        ch_half = cond_dim // 2
+
+        self.cond_in_channels = cond_in_channels
+        self.conv1 = nn.Conv2d(cond_in_channels, ch_half, kernel_size=4, stride=4, padding=0)
+        self.norm1 = _gn(ch_half)
+        self.conv2 = nn.Conv2d(ch_half, ch_half, kernel_size=3, stride=1, padding=1)
+        self.norm2 = _gn(ch_half)
+        self.conv3 = nn.Conv2d(ch_half, cond_dim, kernel_size=4, stride=4, padding=0)
+        self.norm3 = _gn(cond_dim)
+
+        self.resblocks = nn.ModuleList([_ResBlock(cond_dim) for _ in range(n_resblocks)])
+        self.aspp = _ASPP(cond_dim, aspp_dilations) if use_aspp else None
+
+        self.proj = nn.Conv2d(cond_dim, cond_emb_dim, kernel_size=1)
+        self.out_norm = nn.LayerNorm(cond_emb_dim)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        h = F.silu(self.norm1(self.conv1(x)))
+        h = F.silu(self.norm2(self.conv2(h)))
+        h = F.silu(self.norm3(self.conv3(h)))
+        for rb in self.resblocks:
+            h = rb(h)
+        if self.aspp is not None:
+            h = self.aspp(h)
+        h = self.proj(h)
+        b, c, hh, ww = h.shape
+        h = h.view(b, c, hh * ww).permute(0, 2, 1).contiguous()
+        h = self.out_norm(h)
+        return h
+
+
+# ----------------------------------------------------------------------------
+# LLLite module (v2: FiLM + SiLU + 5D path + depth embedding)
+# ----------------------------------------------------------------------------
+
+
+class LLLiteModuleDiT(nn.Module):
+    def __init__(
+        self,
+        name: str,
+        org_module: nn.Linear,
+        cond_emb_dim: int,
+        mlp_dim: int,
+        dropout: float | None = None,
+        multiplier: float = 1.0,
+    ):
+        super().__init__()
+        self.lllite_name = name
+        # Wrap in a list so the original Linear is not registered as a submodule
+        # and its weights stay out of state_dict.
+        self.org_module = [org_module]
+        self.cond_emb_dim = cond_emb_dim
+        self.mlp_dim = mlp_dim
+        self.dropout = dropout
+        self.multiplier = multiplier
+
+        in_dim = org_module.in_features
+
+        self.down = nn.Linear(in_dim, mlp_dim)
+        self.mid = nn.Linear(mlp_dim + cond_emb_dim, mlp_dim)
+
+        # FiLM: cond_local -> (gamma, beta), zero-init for identity at start.
+        self.cond_to_film = nn.Linear(cond_emb_dim, 2 * mlp_dim)
+        nn.init.zeros_(self.cond_to_film.weight)
+        nn.init.zeros_(self.cond_to_film.bias)
+
+        self.up = nn.Linear(mlp_dim, in_dim)
+        nn.init.zeros_(self.up.weight)
+        nn.init.zeros_(self.up.bias)
+
+        self.cond_emb: torch.Tensor | None = None
+        self.org_forward = None
+
+        # Set by the parent ControlNetLLLiteDiT after construction.
+        self.layer_idx: int = -1
+        self._depth_embeds_ref: list[nn.Parameter] = []
+
+    def apply_to(self):
+        if self.org_forward is None:
+            self.org_forward = self.org_module[0].forward
+            self.org_module[0].forward = self.forward
+
+    def restore(self):
+        if self.org_forward is not None:
+            self.org_module[0].forward = self.org_forward
+            self.org_forward = None
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        # Input layouts:
+        #   self/cross attention q/k/v: (B, S, D) — already flattened in the Anima block
+        #   mlp.layer1:                 (B, T, H, W, D) — passed un-flattened
+        # Flatten the 5D case to 3D for the LLLite path and reshape on exit.
+        if self.multiplier == 0.0 or self.cond_emb is None:
+            return self.org_forward(x)
+
+        orig_shape = x.shape
+        is_5d = x.dim() == 5
+        if is_5d:
+            B, T, H, W, D = orig_shape
+            x = x.reshape(B, T * H * W, D)
+
+        cx = self.cond_emb  # (B_c, S, cond_emb_dim)
+
+        # Broadcast cond_emb to the runtime batch (CFG cond+uncond, multi-cond).
+        if x.shape[0] != cx.shape[0]:
+            if x.shape[0] % cx.shape[0] != 0:
+                return self.org_forward(x.reshape(orig_shape) if is_5d else x)
+            cx = cx.repeat(x.shape[0] // cx.shape[0], 1, 1)
+
+        if x.shape[1] != cx.shape[1]:
+            return self.org_forward(x.reshape(orig_shape) if is_5d else x)
+
+        # Run the LLLite mini-MLP in its own parameter dtype, then cast the
+        # correction back to ``x``'s dtype before adding. Robust to autocast
+        # flows where x and LLLite weights have different dtypes.
+        param_dtype = self.down.weight.dtype
+        x_proc = x if x.dtype == param_dtype else x.to(param_dtype)
+        if cx.dtype != param_dtype or cx.device != x.device:
+            cx = cx.to(device=x.device, dtype=param_dtype)
+
+        # Per-module depth embedding (zero-init so it's a no-op at train start).
+        if self._depth_embeds_ref:
+            depth_e = self._depth_embeds_ref[0][self.layer_idx]
+            if depth_e.dtype != param_dtype or depth_e.device != x.device:
+                depth_e = depth_e.to(device=x.device, dtype=param_dtype)
+            cond_local = cx + depth_e
+        else:
+            cond_local = cx
+
+        h = F.silu(self.down(x_proc))
+
+        gb = self.cond_to_film(cond_local)
+        gamma, beta = gb.chunk(2, dim=-1)
+
+        m = self.mid(torch.cat([cond_local, h], dim=-1))
+        m = m * (1 + gamma) + beta
+        m = F.silu(m)
+
+        if self.dropout is not None and self.training:
+            m = F.dropout(m, p=self.dropout)
+
+        out = self.up(m) * self.multiplier
+        if out.dtype != x.dtype:
+            out = out.to(x.dtype)
+
+        y = self.org_forward(x + out)
+
+        if is_5d:
+            # org Linear out_features may differ from in_features — recover with -1.
+            y = y.reshape(orig_shape[0], orig_shape[1], orig_shape[2], orig_shape[3], -1)
+        return y
+
+
+# ----------------------------------------------------------------------------
+# ControlNetLLLiteDiT
+# ----------------------------------------------------------------------------
+
+
+class ControlNetLLLiteDiT(nn.Module):
+    def __init__(
+        self,
+        dit: nn.Module,
+        cond_emb_dim: int = 32,
+        mlp_dim: int = 64,
+        target_layers: str = "self_attn_q",
+        dropout: float | None = None,
+        multiplier: float = 1.0,
+        cond_dim: int = 64,
+        cond_resblocks: int = 1,
+        use_aspp: bool = False,
+        aspp_dilations: tuple[int, ...] = ASPP_DEFAULT_DILATIONS,
+        cond_in_channels: int = 3,
+        inpaint_masked_input: bool = False,
+    ):
+        super().__init__()
+
+        atomics = parse_target_layers(target_layers)
+
+        self.cond_emb_dim = cond_emb_dim
+        self.mlp_dim = mlp_dim
+        self.target_layers = target_layers
+        self.target_atomics = atomics
+        self.dropout = dropout
+        self.multiplier = multiplier
+        self.cond_dim = cond_dim
+        self.cond_resblocks = cond_resblocks
+        self.use_aspp = use_aspp
+        self.aspp_dilations = tuple(aspp_dilations) if use_aspp else ()
+        # 4ch (RGB+mask) inpainting metadata. `inpaint_masked_input` records the training-time
+        # RGB-masking policy for cond_image preparation; it does not alter the forward pass here.
+        self.cond_in_channels = cond_in_channels
+        self.inpaint_masked_input = inpaint_masked_input
+
+        self.conditioning1 = _Conditioning1(
+            cond_dim,
+            cond_emb_dim,
+            cond_resblocks,
+            use_aspp=use_aspp,
+            aspp_dilations=aspp_dilations,
+            cond_in_channels=cond_in_channels,
+        )
+
+        modules = self._create_modules(dit, cond_emb_dim, mlp_dim, atomics, dropout, multiplier)
+        self.lllite_modules = nn.ModuleList(modules)
+
+        n = len(self.lllite_modules)
+        self.depth_embeds = nn.Parameter(torch.zeros(n, cond_emb_dim))
+        for i, m in enumerate(self.lllite_modules):
+            m.layer_idx = i
+            m._depth_embeds_ref = [self.depth_embeds]
+
+        aspp_info = f"aspp={'on' + str(list(self.aspp_dilations)) if use_aspp else 'off'}"
+        inpaint_info = (
+            f", inpaint=on(masked_input={inpaint_masked_input})" if cond_in_channels != 3 else ""
+        )
+        logger.info(
+            "ControlNet-LLLite (Anima v%s): created %d modules for target=%r "
+            "(atomics=%s), cond_in_channels=%d, cond_dim=%d, cond_resblocks=%d, %s, "
+            "cond_emb_dim=%d, mlp_dim=%d%s",
+            LLLITE_ARCH_VERSION,
+            n,
+            target_layers,
+            list(atomics),
+            cond_in_channels,
+            cond_dim,
+            cond_resblocks,
+            aspp_info,
+            cond_emb_dim,
+            mlp_dim,
+            inpaint_info,
+        )
+
+    @staticmethod
+    def _attn_atomic_match(is_self_attn: bool, child_name: str, atomics: tuple[str, ...]) -> bool:
+        if "output_proj" in child_name:
+            return False
+        if is_self_attn:
+            if child_name == "q_proj":
+                return "self_attn_q_pre" in atomics
+            if child_name in ("k_proj", "v_proj"):
+                return "self_attn_kv_pre" in atomics
+            return False
+        else:
+            if child_name == "q_proj":
+                return "cross_attn_q_pre" in atomics
+            return False  # cross_attn K,V live in text-embedding space
+
+    def _create_modules(
+        self,
+        dit: nn.Module,
+        cond_emb_dim: int,
+        mlp_dim: int,
+        atomics: tuple[str, ...],
+        dropout: float | None,
+        multiplier: float,
+    ) -> list[LLLiteModuleDiT]:
+        modules: list[LLLiteModuleDiT] = []
+        want_mlp_fc1 = "mlp_fc1_pre" in atomics
+        any_attn = any(
+            a in atomics for a in ("self_attn_q_pre", "self_attn_kv_pre", "cross_attn_q_pre")
+        )
+
+        for name, module in dit.named_modules():
+            if LLM_ADAPTER_NAME in name:
+                continue
+            cls = module.__class__.__name__
+
+            def _is_linear_like(module):
+                return (
+                hasattr(module, "in_features")
+                and hasattr(module, "out_features")
+                and callable(getattr(module, "forward", None))
+            )
+
+            if any_attn and cls == TARGET_ATTENTION_CLASS:
+                # The Anima-block Attention exposes is_selfattn; the LLM-Adapter
+                # Attention does not — skip the latter even if path filter misses.
+                if not hasattr(module, "is_selfattn"):
+                    continue
+                is_self_attn = bool(module.is_selfattn)
+                for child_name, child in module.named_children():
+                    if not _is_linear_like(child):
+                        continue
+                    if not self._attn_atomic_match(is_self_attn, child_name, atomics):
+                        continue
+                    full_name = f"lllite_dit.{name}.{child_name}".replace(".", "_")
+                    modules.append(
+                        LLLiteModuleDiT(
+                            full_name, child, cond_emb_dim, mlp_dim, dropout, multiplier
+                        )
+                    )
+
+            elif want_mlp_fc1 and cls == TARGET_MLP_CLASS:
+                child = getattr(module, "layer1", None)
+                if not _is_linear_like(child):
+                    continue
+                full_name = f"lllite_dit.{name}.layer1".replace(".", "_")
+                modules.append(
+                    LLLiteModuleDiT(full_name, child, cond_emb_dim, mlp_dim, dropout, multiplier)
+                )
+
+        return modules
+
+    def set_cond_image(self, cond_image: torch.Tensor | None):
+        """cond_image: (B, 3, H*16, W*16) in [-1, 1]; ``None`` clears."""
+        if cond_image is None:
+            for m in self.lllite_modules:
+                m.cond_emb = None
+            return
+        cx = self.conditioning1(cond_image)  # (B, S, cond_emb_dim)
+        for m in self.lllite_modules:
+            m.cond_emb = cx
+
+    def clear_cond_image(self):
+        self.set_cond_image(None)
+
+    def set_multiplier(self, multiplier: float):
+        self.multiplier = multiplier
+        for m in self.lllite_modules:
+            m.multiplier = multiplier
+
+    def apply_to(self):
+        for m in self.lllite_modules:
+            m.apply_to()
+
+    def restore(self):
+        for m in self.lllite_modules:
+            m.restore()
+
+
+# ----------------------------------------------------------------------------
+# Save / load (named-key format; legacy lllite_modules.* is rejected)
+# ----------------------------------------------------------------------------
+
+_INTERNAL_MODULES_PREFIX = "lllite_modules."
+_INTERNAL_COND_PREFIX = "conditioning1."
+_INTERNAL_DEPTH_KEY = "depth_embeds"
+_SAVED_COND_PREFIX = "lllite_conditioning1."
+_SAVED_DEPTH_SUFFIX = ".depth_embed"
+
+
+def _from_saved_state_dict(lllite: ControlNetLLLiteDiT, weights_sd: dict) -> dict:
+    """Rewrite a v2 named-key state dict back to the internal layout."""
+    name_to_idx = {m.lllite_name: i for i, m in enumerate(lllite.lllite_modules)}
+    n_modules = len(name_to_idx)
+    out: dict = {}
+    depth_slices: dict = {}
+
+    for k, v in weights_sd.items():
+        if k.startswith(_SAVED_COND_PREFIX):
+            out[_INTERNAL_COND_PREFIX + k[len(_SAVED_COND_PREFIX) :]] = v
+            continue
+        if k.endswith(_SAVED_DEPTH_SUFFIX):
+            name = k[: -len(_SAVED_DEPTH_SUFFIX)]
+            if name in name_to_idx:
+                depth_slices[name_to_idx[name]] = v
+                continue
+        head, dot, tail = k.partition(".")
+        if dot and head in name_to_idx:
+            out[f"{_INTERNAL_MODULES_PREFIX}{name_to_idx[head]}.{tail}"] = v
+            continue
+        out[k] = v
+
+    if depth_slices:
+        missing = [i for i in range(n_modules) if i not in depth_slices]
+        if missing:
+            raise RuntimeError(f"depth_embed slices missing for module idx(es) {missing}")
+        out[_INTERNAL_DEPTH_KEY] = torch.stack([depth_slices[i] for i in range(n_modules)], dim=0)
+
+    return out
+
+
+def load_lllite_weights(lllite: ControlNetLLLiteDiT, file: str, strict: bool = False):
+    weights_sd = safetensors.torch.load_file(file)
+
+    if any(k.startswith(_INTERNAL_MODULES_PREFIX) for k in weights_sd):
+        raise RuntimeError(
+            f"weights at {file} appear to be in a legacy ControlNet-LLLite weight format "
+            f"(keys starting with '{_INTERNAL_MODULES_PREFIX}'). The current code uses a "
+            f"named-key format (per-module key prefix = lllite_name, e.g. "
+            f"'lllite_dit_blocks_0_self_attn_q_proj.down.weight'). Re-train with the current codebase."
+        )
+
+    converted = _from_saved_state_dict(lllite, weights_sd)
+    info = lllite.load_state_dict(converted, strict=strict)
+    logger.info("loaded LLLite weights from %s: %s", file, info)
+    return info
+
+
+def read_lllite_metadata(file: str) -> dict:
+    if os.path.splitext(file)[1] != ".safetensors":
+        raise RuntimeError(f"Must use .safetensors files, got {file}")
+
+    with safetensors.safe_open(file, framework="pt") as f:
+        return f.metadata() or {}
+
+
+# ----------------------------------------------------------------------------
+# ComfyUI nodes for Anima ControlNet-LLLite
+# ----------------------------------------------------------------------------
+
+
+def _get_inner_dit(model) -> torch.nn.Module:
+    """Reach the underlying Anima DiT (nn.Module) from a ComfyUI ModelPatcher."""
+    inner = getattr(model, "model", None)
+    if inner is None:
+        raise RuntimeError("Input MODEL has no .model attribute (not a ModelPatcher?)")
+    dit = getattr(inner, "diffusion_model", None)
+    if dit is None:
+        raise RuntimeError("MODEL.model has no .diffusion_model — not a UNet/DiT model?")
+    return dit
+
+
+def _target_cond_hw(latent_h: int, latent_w: int, patch_spatial: int = 2) -> tuple[int, int]:
+    """Return the (H, W) the cond image / mask must be resized to.
+
+    The LLLite ``conditioning1`` Conv has stride 16, so the cond image must be
+    sized to ``latent_HW * 8`` in input pixel space (= ``token_HW * 16`` after
+    DiT patchify with patch_spatial=2). The DiT internally pads the latent up
+    to a multiple of ``patch_spatial`` (see ``MiniTrainDIT.forward`` →
+    ``pad_to_patch_size``), so we mirror that rounding here — otherwise odd
+    latent dims (e.g. 1032 px → 129 latent) yield a token-count mismatch that
+    silently bypasses every LLLite module.
+    """
+    padded_h = ((latent_h + patch_spatial - 1) // patch_spatial) * patch_spatial
+    padded_w = ((latent_w + patch_spatial - 1) // patch_spatial) * patch_spatial
+    return padded_h * 8, padded_w * 8
+
+
+def _prepare_cond_image(
+    image: torch.Tensor,
+    latent_h: int,
+    latent_w: int,
+    device: torch.device,
+    dtype: torch.dtype,
+    patch_spatial: int = 2,
+) -> torch.Tensor:
+    """ComfyUI IMAGE (B,H,W,3) in [0,1] → (1,3,H*8,W*8) in [-1,1]."""
+    if image.ndim == 4 and image.shape[-1] == 3:
+        # (B, H, W, 3) -> (B, 3, H, W)
+        img = image.permute(0, 3, 1, 2).contiguous()
+    else:
+        raise ValueError(f"Unexpected cond image shape: {tuple(image.shape)} (expected B,H,W,3)")
+
+    img = img[:1]  # use first frame only
+    target_h, target_w = _target_cond_hw(latent_h, latent_w, patch_spatial)
+    if img.shape[-2] != target_h or img.shape[-1] != target_w:
+        img = F.interpolate(img, size=(target_h, target_w), mode="bicubic", align_corners=False)
+        img = img.clamp(0.0, 1.0)
+    img = img * 2.0 - 1.0
+    return img.to(device=device, dtype=dtype)
+
+
+def _prepare_mask(
+    mask: torch.Tensor,
+    latent_h: int,
+    latent_w: int,
+    device: torch.device,
+    dtype: torch.dtype,
+    patch_spatial: int = 2,
+) -> torch.Tensor:
+    """ComfyUI MASK (B,H,W) in [0,1] → (1,1,H*8,W*8) binarized at 0.5.
+
+    Returns the mask in ``{0.0, 1.0}`` (1 = inpaint area, 0 = keep). The caller
+    is responsible for the ``*2-1`` rescale before concat with RGB.
+    """
+    if mask.ndim == 3:
+        m = mask.unsqueeze(1)  # (B, 1, H, W)
+    elif mask.ndim == 4 and mask.shape[1] == 1:
+        m = mask
+    else:
+        raise ValueError(f"Unexpected mask shape: {tuple(mask.shape)} (expected B,H,W or B,1,H,W)")
+
+    m = m[:1]
+    target_h, target_w = _target_cond_hw(latent_h, latent_w, patch_spatial)
+    if m.shape[-2] != target_h or m.shape[-1] != target_w:
+        m = F.interpolate(m.float(), size=(target_h, target_w), mode="nearest")
+    m = (m >= 0.5).to(dtype=dtype)
+    return m.to(device=device)
+
+
+def _build_inpaint_cond_image(
+    rgb_pm1: torch.Tensor, mask01: torch.Tensor, masked_input: bool
+) -> torch.Tensor:
+    """rgb_pm1: (1,3,H,W) in [-1,1], mask01: (1,1,H,W) in {0,1}. Returns (1,4,H,W).
+
+    Mirrors ``_build_inpaint_cond_image`` in the sd-scripts training / inference
+    code: the mask channel is rescaled to ``[-1, +1]`` (matches the RGB range),
+    and if ``masked_input`` is set the RGB is zeroed where ``mask >= 0.5``.
+    """
+    if masked_input:
+        keep = (mask01 < 0.5).to(rgb_pm1.dtype)
+        rgb_pm1 = rgb_pm1 * keep
+    mask_pm1 = mask01.to(rgb_pm1.dtype) * 2.0 - 1.0
+    return torch.cat([rgb_pm1, mask_pm1], dim=1)
+
+
+ETNControlNet = io.Custom("ETN_CONTROL_NET")
+
+
+class ControlLoad(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_control_load",
+            display_name="Load ControlNet (tooling-nodes)",
+            description="Loads ControlNet weights. Currently only supports Anima LLLite weights.",
+            category="external_tooling",
+            inputs=[
+                io.Model.Input("model"),
+                io.Combo.Input("weights", folder_paths.get_filename_list("controlnet")),
+            ],
+            outputs=[
+                io.Model.Output("out_model", "model"),
+                ETNControlNet.Output("control_net"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, model: ModelPatcher, weights: str):  # type: ignore[override]
+        weights_path = folder_paths.get_full_path("controlnet", weights)
+        if weights_path is None or not os.path.isfile(weights_path):
+            raise FileNotFoundError(f"LLLite weights not found: {weights}")
+
+        # Architecture is fully determined by the trained weights — read everything
+        # from metadata rather than exposing knobs that would just cause load errors.
+        meta = read_lllite_metadata(weights_path)
+        if "lllite.version" not in meta:
+            raise RuntimeError(
+                "Unrecognized model. This node currently only loads Anima LLLite weights."
+            )
+        ce_dim = int(meta.get("lllite.cond_emb_dim", 32))
+        m_dim = int(meta.get("lllite.mlp_dim", 64))
+        # v2 records the canonical atomic form under lllite.target_atomics; fall back
+        # to the legacy preset key, then to the v1 default.
+        tl = meta.get("lllite.target_atomics", meta.get("lllite.target_layers", "self_attn_q"))
+        cond_dim = int(meta.get("lllite.cond_dim", 64))
+        cond_resblocks = int(meta.get("lllite.cond_resblocks", 1))
+        use_aspp = str(meta.get("lllite.use_aspp", "false")).lower() == "true"
+        aspp_dilations_meta = meta.get("lllite.aspp_dilations")
+        if use_aspp and aspp_dilations_meta:
+            aspp_dilations = tuple(int(d) for d in aspp_dilations_meta.split(",") if d.strip())
+        else:
+            aspp_dilations = ASPP_DEFAULT_DILATIONS
+        cond_in_channels = int(meta.get("lllite.cond_in_channels", 3))
+        inpaint_masked_input = (
+            str(meta.get("lllite.inpaint_masked_input", "false")).lower() == "true"
+        )
+
+        lllite = ControlNetLLLiteDiT(
+            _get_inner_dit(model),
+            cond_emb_dim=ce_dim,
+            mlp_dim=m_dim,
+            target_layers=tl,
+            multiplier=1.0,
+            cond_dim=cond_dim,
+            cond_resblocks=cond_resblocks,
+            use_aspp=use_aspp,
+            aspp_dilations=aspp_dilations,
+            cond_in_channels=cond_in_channels,
+            inpaint_masked_input=inpaint_masked_input,
+        )
+        load_lllite_weights(lllite, weights_path, strict=False)
+        lllite.eval().requires_grad_(False)
+        return io.NodeOutput(model, lllite)
+
+
+class ControlApply(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_control_apply",
+            display_name="Apply ControlNet (tooling-nodes)",
+            description="Applies ControlNet conditioning. Currently only supports Anima LLLite weights.",
+            category="external_tooling",
+            inputs=[
+                io.Model.Input("model"),
+                ETNControlNet.Input("control_net"),
+                io.Image.Input("image"),
+                io.Mask.Input("mask", optional=True),
+                io.Float.Input("strength", default=1.0, min=-10.0, max=10.0, step=0.01),
+                io.Float.Input("start_percent", default=0.0, min=0.0, max=1.0, step=0.001),
+                io.Float.Input("end_percent", default=1.0, min=0.0, max=1.0, step=0.001),
+            ],
+            outputs=[io.Model.Output("model")],
+        )
+
+    @classmethod
+    def execute(  # type: ignore[override]
+        cls,
+        model: ModelPatcher,
+        control_net: ControlNetLLLiteDiT,
+        image: torch.Tensor,
+        strength: float,
+        start_percent: float,
+        end_percent: float,
+        mask: torch.Tensor | None = None,
+    ):
+        dit = _get_inner_dit(model)
+        patch_spatial = int(getattr(dit, "patch_spatial", 2))
+
+        lllite = control_net
+        lllite.set_multiplier(strength)
+
+        # Mask / cond_in_channels consistency: 4ch weights need a MASK, 3ch weights ignore it.
+        if lllite.cond_in_channels == 4 and mask is None:
+            raise ValueError("ControlNet weights require a mask input (inpaint mode)")
+        if lllite.cond_in_channels != 4 and mask is not None:
+            mask = None
+
+        # Convert percent range -> sigma range (start_percent=0 → sigma_max).
+        model_sampling = model.get_model_object("model_sampling")
+        sigma_start = float(model_sampling.percent_to_sigma(start_percent))
+        sigma_end = float(model_sampling.percent_to_sigma(end_percent))
+
+        # Capture image / mask tensors (cloned to detach from any upstream caching)
+        src_image = image.detach().clone()
+        src_mask = mask.detach().clone() if mask is not None else None
+        is_inpaint = lllite.cond_in_channels == 4
+
+        # Cache for the per-resolution preprocessed cond image (avoids repeat resize)
+        cache: dict[str, Any] = {"cond_image_pp": None, "key": None, "lllite_loaded_to": None}
+
+        # Capture any previously-installed wrapper BEFORE we clone — model_options
+        # has a single "model_function_wrapper" slot, so without delegation a second
+        # wrapper-installing node would silently no-op the first. Mirrors the
+        # ChromaRadianceOptions pattern in comfy_extras/nodes_chroma_radiance.py.
+        old_wrapper = model.model_options.get("model_function_wrapper")
+
+        def _call_next(apply_model, input_x, timestep, c):
+            if old_wrapper is not None:
+                return old_wrapper(apply_model, {"input": input_x, "timestep": timestep, "c": c})
+            return apply_model(input_x, timestep, **c)
+
+        def wrapper(apply_model, args):
+            input_x = args["input"]
+            timestep = args["timestep"]
+            c = args["c"]
+
+            # Step-range gate: skip LLLite entirely when current sigma is outside
+            # [sigma_end, sigma_start]. percent_to_sigma maps 0.0 → sigma_max,
+            # 1.0 → sigma_min, so the active window is sigma_end <= sigma <= sigma_start.
+            sigma = float(timestep.max().item())
+            if not (sigma_end <= sigma <= sigma_start):
+                return _call_next(apply_model, input_x, timestep, c)
+
+            # Anima latent shape: (B, C, T, H, W) — take spatial dims from the tail.
+            latent_h, latent_w = int(input_x.shape[-2]), int(input_x.shape[-1])
+            device = input_x.device
+            dtype = input_x.dtype
+
+            # Move LLLite to the runtime device/dtype lazily.
+            tag = (device, dtype)
+            if cache["lllite_loaded_to"] != tag:
+                lllite.to(device=device, dtype=dtype)
+                cache["lllite_loaded_to"] = tag
+                cache["cond_image_pp"] = None  # invalidate
+
+            key = (latent_h, latent_w, device, dtype)
+            if cache["key"] != key or cache["cond_image_pp"] is None:
+                rgb = _prepare_cond_image(
+                    src_image, latent_h, latent_w, device, dtype, patch_spatial
+                )
+                if is_inpaint:
+                    assert src_mask is not None, "Cannot use inpaint control-net without a mask"
+                    mk = _prepare_mask(src_mask, latent_h, latent_w, device, dtype, patch_spatial)
+                    cache["cond_image_pp"] = _build_inpaint_cond_image(
+                        rgb, mk, lllite.inpaint_masked_input
+                    )
+                else:
+                    cache["cond_image_pp"] = rgb
+                cache["key"] = key
+
+            lllite.set_multiplier(strength)
+            lllite.set_cond_image(cache["cond_image_pp"])
+            lllite.apply_to()
+            try:
+                return _call_next(apply_model, input_x, timestep, c)
+            finally:
+                lllite.restore()
+                lllite.clear_cond_image()
+
+        m = model.clone()
+        m.set_model_unet_function_wrapper(wrapper)
+        return (m,)

js/external_tooling_nodes.js

@@ -32,7 +32,6 @@ function loadImage(base64) {
 }
 
 const canvasIcon = loadImage("data:image/webp;base64,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")
-const outputIcon = loadImage("data:image/webp;base64,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")
 
 function setIconImage(nodeType, image, size, padRows, padCols) {
     const onAdded = nodeType.prototype.onAdded
@@ -72,36 +71,70 @@ const parameterTypes = {
     "text": ["text", "prompt (positive)", "prompt (negative)"],
 }
 
-function changeWidget(widget, type, value, options) {
+function defaultParameterType(widgetType, connectedNode, connectedWidget) {
-    widget.type = type
+    let paramType = parameterTypes[widgetType][0]
-    widget.value = value
+    if (connectedNode.comfyClass === "CLIPTextEncode") {
-    widget.options = options
+        paramType = "prompt (positive)"
+    }
+    const round = connectedWidget.options?.round
+    if ((paramType == "number" && round === undefined) || round === 1) {
+        paramType = "number (integer)"
+    }
+    return paramType
 }
 
-function changeWidgets(node, type, value, options) {
+function valueMatchesType(value, type, options) {
+    if (type === "number") {
+        return typeof value === "number"
+    } else if (type === "combo") {
+        return options?.values?.includes(value)
+    } else if (type === "toggle") {
+        return typeof value === "boolean"
+    }
+    return typeof value === "string"
+}
+
+function optionalWidgetValue(widgets, index, fallback) {
+    const result = widgets.length > index ? widgets[index].value : null
+    return result === null || result === -1e10 || result === 1e10 ? fallback : result
+}
+
+function changeWidgets(node, type, connectedNode, connectedWidget) {
     if (type === "customtext") {
         type = "text"
     }
+    const options = connectedWidget.options
 
-    node.widgets[1].value = parameterTypes[type][0]
+    const parameterTypeHint = node.widgets[1].value
-    node.widgets[1].options = {values: parameterTypes[type]}
+    const notSpecialized = node.widgets[1].options.values.includes("auto")
-    changeWidget(node.widgets[2], type, value, options)
+    const parameterTypeMismatch = !parameterTypes[type].includes(parameterTypeHint)
-
+    if (notSpecialized || parameterTypeMismatch) {
+        node.widgets[1].options = {values: parameterTypes[type]}
+    }
+    if (parameterTypeMismatch) {
+        node.widgets[1].value = defaultParameterType(type, connectedNode, connectedWidget)
+    }
+    const oldDefault = node.widgets.length > 2 ? node.widgets[2].value : connectedWidget.value
+    const oldMin = optionalWidgetValue(node.widgets, 3, options?.min ?? 0)
+    const oldMax = optionalWidgetValue(node.widgets, 4, options?.max ?? 100)
+    const isDefaultValid = valueMatchesType(oldDefault, type, connectedWidget.options)
+    while (node.widgets.length > 2) {
+        node.widgets.pop()
+    }
+    const value = isDefaultValid && oldDefault !== "" ? oldDefault : connectedWidget.value
+    node.addWidget(type, "default", value, null, options)
     if (type === "number") {
-        changeWidget(node.widgets[3], "number", options?.min ?? 0, options)
+        node.addWidget("number", "min", oldMin, null, options)
-        changeWidget(node.widgets[4], "number", options?.max ?? 100, options)
+        node.addWidget("number", "max", oldMax, null, options)
-    } else {
-        changeWidget(node.widgets[3], "number", 0, {min: 0, max: 0})
-        changeWidget(node.widgets[4], "number", 0, {min: 0, max: 0})
     }
 }
 
 function adaptWidgetsToConnection(node) {
-    if (!node.outputs || node.outputs.length === 0 || !node.outputs[0].links) {
+    if (!node.outputs || node.outputs.length === 0) {
         return
     }
     const links = node.outputs[0].links
-    if (links.length === 1) {
+    if (links && links.length === 1) {
         const link = node.graph.links[links[0]]
         if (!link) return
 
@@ -109,7 +142,7 @@ function adaptWidgetsToConnection(node) {
         if (!theirNode || !theirNode.inputs) return
     
         const input = theirNode.inputs[link.target_slot]
-        if (!input) return
+        if (!input || !input.widget || theirNode.widgets === undefined) return
 
         node.outputs[0].type = input.type
 
@@ -119,11 +152,15 @@ function adaptWidgetsToConnection(node) {
 
         const widgetName = input.widget.name
         const theirWidget = theirNode.widgets.find((w) => w.name === widgetName)
-        const widgetType = theirWidget.origType ?? theirWidget.type
+        if (!theirWidget) return // connected to a custom node that isn't installed
-        changeWidgets(node, widgetType, theirWidget.value, theirWidget.options)
 
-    } else if (links.length === 0) {
+        const widgetType = theirWidget.origType ?? theirWidget.type
+        changeWidgets(node, widgetType, theirNode, theirWidget)
+
+    } else if (!links || links.length === 0) {
         node.outputs[0].type = "*"
+        node.widgets[1].value = "auto"
+        node.widgets[1].options = {values: ["auto"]}
     }
 }
 
@@ -169,10 +206,14 @@ app.registerExtension({
     beforeRegisterNodeDef(nodeType /*typeof LGraphNode*/, nodeData /*ComfyObjectInfo*/, app) {
         if (nodeData.name === "ETN_KritaCanvas") {
             setIconImage(nodeType, canvasIcon, [200, 100], 0, 2)
-        } else if (nodeData.name === "ETN_KritaOutput") {
+        } else if (nodeData.name === "ETN_Parameter") {
-            setIconImage(nodeType, outputIcon, [200, 120], 2, 0)
-        } else if (nodeData.name == "ETN_Parameter") {
             setupParameterNode(nodeType)
+        } else if (nodeData.name === "ETN_SendText") {
+            const onAdded = nodeType.prototype.onAdded
+            nodeType.prototype.onAdded = function() {
+                onAdded?.apply(this, arguments)
+                this.inputs[0].type = "*"
+            }
         }
     },
 

krita.py

@@ -1,11 +1,16 @@
-import torch
+import sys
-import numpy as np
+from enum import Enum
 from pathlib import Path
-from typing import NamedTuple
+from typing import Any, NamedTuple
+
+import comfy.samplers
+import numpy as np
+import server
+import torch
+from comfy.comfy_types.node_typing import IO
+from comfy_api.latest import io
 from PIL import Image
 
-import server
-import comfy.samplers
 from .nodes import SendImageWebSocket
 
 
@@ -34,8 +39,11 @@ class WorkflowExchange:
         for publisher in self._publishers.values():
             await self._notify(client_id, publisher)
 
-    def unsubscribe(self, client_id: str):
+    async def unsubscribe(self, client_id: str):
-        self._subscribers.remove(client_id)
+        if client_id in self._subscribers:
+            self._subscribers.remove(client_id)
+        else:
+            raise KeyError("No subscriber found with id " + client_id)
 
     async def _notify(self, client_id: str, publisher: Publisher):
         data = {
@@ -52,75 +60,204 @@ def _placeholder_image():
     return torch.from_numpy(image)[None,]
 
 
-class KritaOutput(SendImageWebSocket):
+class _BasicTypes(str):
-    RETURN_TYPES = ()
+    """Matches IO.PRIMITIVE, but also any list of choices"""
-    FUNCTION = "send_images"
+
-    OUTPUT_NODE = True
+    basic_types = IO.PRIMITIVE.split(",")  # STRING, FLOAT, INT, BOOLEAN
-    CATEGORY = "krita"
+
+    def __eq__(self, other):
+        return other in self.basic_types or isinstance(other, (list, _BasicTypes))
+
+    def __ne__(self, other):
+        return not self.__eq__(other)
 
 
-class KritaCanvas:
+BasicTypes = _BasicTypes("BASIC")
+
+
+class OutputBatchMode(Enum):
+    default = "default"
+    images = "images"
+    animation = "animation"
+    layers = "layers"
+
+
+class KritaOutput(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {}
+        return io.Schema(
+            node_id="ETN_KritaOutput",
+            display_name="Krita Output",
+            category="krita",
+            inputs=[
+                io.Image.Input("images"),
+                io.Int.Input("x", "offset x", default=0),
+                io.Int.Input("y", "offset y", default=0),
+                io.String.Input("name", default=""),
+                io.Combo.Input(
+                    "batch_mode", OutputBatchMode, "batch mode", default=OutputBatchMode.default
+                ),
+                io.Boolean.Input("resize_canvas", "resize canvas", default=False),
+            ],
+            is_output_node=True,
+        )
 
-    RETURN_TYPES = ("IMAGE", "INT", "INT", "INT")
-    RETURN_NAMES = ("image", "width", "height", "seed")
-    FUNCTION = "placeholder"
-    CATEGORY = "krita"
-
-    def placeholder(self):
-        return (_placeholder_image(), 512, 512, 0)
-
-
-class KritaSelection:
     @classmethod
-    def INPUT_TYPES(cls):
+    def execute(  # type: ignore
-        return {}
+        cls,
-
+        images: torch.Tensor,
-    RETURN_TYPES = ("MASK",)
+        x: int = 0,
-    RETURN_NAMES = ("mask",)
+        y: int = 0,
-    FUNCTION = "placeholder"
+        name="",
-    CATEGORY = "krita"
+        batch_mode: OutputBatchMode | str = OutputBatchMode.default,
-
+        resize_canvas=False,
-    def placeholder(self):
+    ):
-        return (torch.ones(1, 512, 512),)
+        batch_mode = batch_mode.value if isinstance(batch_mode, OutputBatchMode) else batch_mode
-
+        info = {
-
+            "name": name,
-class KritaImageLayer:
+            "offset_x": x,
-    @classmethod
+            "offset_y": y,
-    def INPUT_TYPES(cls):
+            "batch_mode": batch_mode,
-        return {
+            "resize_canvas": resize_canvas,
-            "required": {
-                "name": ("STRING", {"default": "Image"}),
-            }
         }
-
+        output = SendImageWebSocket.execute(images, "PNG")
-    RETURN_TYPES = ("IMAGE",)
+        assert isinstance(output.ui, dict)
-    RETURN_NAMES = ("image",)
+        output.ui["info"] = [info]
-    FUNCTION = "placeholder"
+        return output
-    CATEGORY = "krita"
-
-    def placeholder(self, name: str):
-        return (_placeholder_image(),)
 
 
-class KritaMaskLayer:
+class KritaSendText(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {
+        return io.Schema(
-            "required": {
+            node_id="ETN_KritaSendText",
-                "name": ("STRING", {"default": "Mask"}),
+            display_name="Send Text",
-            }
+            category="krita",
-        }
+            inputs=[
+                io.AnyType.Input("value"),
+                io.String.Input("name", default="Output"),
+                io.Combo.Input("type", options=["text", "markdown", "html"], default="text"),
+            ],
+            is_output_node=True,
+        )
 
-    RETURN_TYPES = ("MASK",)
+    @classmethod
-    RETURN_NAMES = ("mask",)
+    def execute(cls, value: Any, name: str, type: str):  # type: ignore
-    FUNCTION = "placeholder"
+        mime = {
-    CATEGORY = "krita"
+            "text": "text/plain",
+            "markdown": "text/markdown",
+            "html": "text/html",
+        }[type]
+        text = "None"
+        if value is not None:
+            try:
+                text = str(value)
+            except Exception as e:
+                text = f"Could not convert to text: {e}"
 
-    def placeholder(self, name: str):
+        return io.NodeOutput(ui={"text": [{"name": name, "text": text, "content-type": mime}]})
-        return (torch.ones(1, 512, 512),)
+
+
+class KritaCanvas(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_KritaCanvas",
+            display_name="Krita Canvas",
+            category="krita",
+            outputs=[
+                io.Image.Output(display_name="image"),
+                io.Int.Output(display_name="width"),
+                io.Int.Output(display_name="height"),
+                io.Int.Output(display_name="seed"),
+                io.Mask.Output(display_name="mask"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, **kwargs):
+        return io.NodeOutput(_placeholder_image(), 512, 512, 0, torch.ones(1, 512, 512))
+
+
+class SelectionContext(Enum):
+    automatic = "automatic"
+    entire_image = "entire image"
+    mask_bounds = "mask bounds"
+
+
+_selection_context_help = """
+Determines the section (crop bounding box) of the image and mask to transmit:
+- automatic: area around the selection determined by Krita settings
+- entire image: always use the entire canvas area
+- mask bounds: tight bounding box of the current selection
+
+This affects the Selection and Canvas nodes. The offset x/y outputs indicate the top-left corner of the context area relative to the full canvas."""
+
+
+class KritaSelection(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_KritaSelection",
+            display_name="Krita Selection",
+            category="krita",
+            inputs=[
+                io.Combo.Input(
+                    "context",
+                    options=SelectionContext,
+                    default=SelectionContext.entire_image,
+                    tooltip=_selection_context_help,
+                ),
+                io.Int.Input("padding", "padding", default=0, min=0),
+            ],
+            outputs=[
+                io.Mask.Output("mask", "mask"),
+                io.Boolean.Output("active", "active"),
+                io.Int.Output("x", "offset x"),
+                io.Int.Output("y", "offset y"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, **kwargs):
+        return io.NodeOutput(torch.ones(1, 512, 512), False, 0, 0)
+
+
+class KritaImageLayer(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_KritaImageLayer",
+            display_name="Krita Image Layer",
+            category="krita",
+            inputs=[io.String.Input("name", default="Image")],
+            outputs=[
+                io.Image.Output(display_name="image"),
+                io.Mask.Output(display_name="mask"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, name: str):  # type: ignore
+        return io.NodeOutput(_placeholder_image(), torch.ones(1, 512, 512))
+
+
+class KritaMaskLayer(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_KritaMaskLayer",
+            display_name="Krita Mask Layer",
+            category="krita",
+            inputs=[io.String.Input("name", default="Mask")],
+            outputs=[
+                io.Mask.Output(display_name="mask"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, name: str):  # type: ignore
+        return io.NodeOutput(torch.ones(1, 512, 512))
 
 
 _param_types = [
@@ -133,64 +270,95 @@ _param_types = [
     "prompt (positive)",
     "prompt (negative)",
 ]
+_fmax = sys.float_info.max
 
 
-class Parameter:
+class Parameter(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {
+        return io.Schema(
-            "required": {
+            node_id="ETN_Parameter",
-                "name": ("STRING", {"default": "Parameter"}),
+            display_name="Parameter",
-                "type": (_param_types, {"default": "auto"}),
+            category="krita",
-                "default": ("STRING", {"default": ""}),
+            inputs=[
-                "min": ("FLOAT", {"default": 0.0}),
+                io.String.Input("name", default="Parameter"),
-                "max": ("FLOAT", {"default": 1.0}),
+                io.Combo.Input("type", options=_param_types, default="auto"),
-            }
+                io.String.Input("default", default=""),
-        }
+                io.Float.Input("min", default=-1e10, min=-_fmax, max=_fmax, optional=True),
+                io.Float.Input("max", default=1e10, min=-_fmax, max=_fmax, optional=True),
+            ],
+            outputs=[io.AnyType.Output(display_name="value")],
+        )
 
-    RETURN_TYPES = ("*",)
-    RETURN_NAMES = ("value",)
-    FUNCTION = "placeholder"
-    CATEGORY = "krita"
-
-    def placeholder(self, name: str, type: str, default, min, max):
-        return (default,)
-
-
-class KritaStyle:
     @classmethod
-    def INPUT_TYPES(cls):
+    def execute(cls, name: str, type: str, default, min=0.0, max=1.0):  # type: ignore
-        return {
+        if type == "number":
-            "required": {
+            return io.NodeOutput(float(default))
-                "name": ("STRING", {"default": "Style"}),
+        elif type == "number (integer)":
-                "sampler_preset": (["auto", "regular", "live"],),
+            return io.NodeOutput(int(default))
-            }
+        return io.NodeOutput(default)
-        }
 
-    RETURN_TYPES = (
-        "MODEL",
-        "CLIP",
-        "VAE",
-        "STRING",
-        "STRING",
-        comfy.samplers.KSampler.SAMPLERS,
-        comfy.samplers.KSampler.SCHEDULERS,
-        "INT",
-        "FLOAT",
-    )
-    RETURN_NAMES = (
-        "model",
-        "clip",
-        "vae",
-        "positive prompt",
-        "negative prompt",
-        "sampler name",
-        "scheduler",
-        "steps",
-        "guidance",
-    )
-    FUNCTION = "placeholder"
-    CATEGORY = "krita"
 
-    def placeholder(self, name: str, sampler_preset: str):
+class KritaStyle(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_KritaStyle",
+            display_name="Krita Style",
+            category="krita",
+            inputs=[
+                io.String.Input("name", default="Style"),
+                io.Combo.Input("sampler_preset", options=["auto", "regular", "live"]),
+            ],
+            outputs=[
+                io.Model.Output(display_name="model"),
+                io.Clip.Output(display_name="clip"),
+                io.Vae.Output(display_name="vae"),
+                io.String.Output(display_name="positive prompt"),
+                io.String.Output(display_name="negative prompt"),
+                io.Combo.Output(
+                    display_name="sampler name", options=comfy.samplers.KSampler.SAMPLERS
+                ),
+                io.Combo.Output(
+                    display_name="scheduler", options=comfy.samplers.KSampler.SCHEDULERS
+                ),
+                io.Int.Output(display_name="steps"),
+                io.Float.Output(display_name="guidance"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, name: str, sampler_preset: str):  # type: ignore
+        raise NotImplementedError("This workflow must be started from Krita!")
+
+
+class KritaStyleAndPrompt(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_KritaStyleAndPrompt",
+            display_name="Krita Style & Prompt",
+            category="krita",
+            inputs=[
+                io.Combo.Input("sampler_preset", options=["auto", "regular", "live"]),
+            ],
+            outputs=[
+                io.Model.Output(display_name="model (with loras)"),
+                io.Clip.Output(display_name="clip"),
+                io.Vae.Output(display_name="vae"),
+                io.String.Output(display_name="positive prompt (evaluated)"),
+                io.String.Output(display_name="negative prompt (evaluated)"),
+                io.Combo.Output(
+                    display_name="sampler name", options=comfy.samplers.KSampler.SAMPLERS
+                ),
+                io.Combo.Output(
+                    display_name="scheduler", options=comfy.samplers.KSampler.SCHEDULERS
+                ),
+                io.Int.Output(display_name="steps"),
+                io.Float.Output(display_name="guidance"),
+            ],
+        )
+
+    @classmethod
+    def execute(cls, name: str, sampler_preset: str):  # type: ignore
         raise NotImplementedError("This workflow must be started from Krita!")

nodes.py

@@ -1,30 +1,45 @@
 from __future__ import annotations
-from PIL import Image
+
-import numpy as np
 import base64
-import torch
+import time
+from copy import copy
+from dataclasses import dataclass
 from io import BytesIO
-from server import PromptServer, BinaryEventTypes
+from typing import NamedTuple
+from uuid import uuid4
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+from comfy.clip_vision import ClipVisionModel
+from comfy.sd import StyleModel
+from comfy_api.latest import io
+from PIL import Image
+from server import BinaryEventTypes, PromptServer
 
 
-class LoadImageBase64:
+class LoadImageBase64(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(s):
+    def define_schema(cls):
-        return {"required": {"image": ("STRING", {"multiline": False})}}
+        return io.Schema(
+            node_id="ETN_LoadImageBase64",
+            display_name="Load Image (Base64)",
+            category="external_tooling",
+            inputs=[io.String.Input("image", multiline=False)],
+            outputs=[io.Image.Output(display_name="image"), io.Mask.Output(display_name="mask")],
+        )
 
-    RETURN_TYPES = ("IMAGE", "MASK")
+    @classmethod
-    CATEGORY = "external_tooling"
+    def execute(cls, image: str):  # type: ignore
-    FUNCTION = "load_image"
+        _strip_prefix(image, "data:image/png;base64,")
-
-    def load_image(self, image):
         imgdata = base64.b64decode(image)
         img = Image.open(BytesIO(imgdata))
 
         if "A" in img.getbands():
             mask = np.array(img.getchannel("A")).astype(np.float32) / 255.0
-            mask = 1.0 - torch.from_numpy(mask)
+            mask = torch.from_numpy(mask)[None,]
         else:
-            mask = torch.zeros((64, 64), dtype=torch.float32, device="cpu")
+            mask = None
 
         img = img.convert("RGB")
         img = np.array(img).astype(np.float32) / 255.0
@@ -33,16 +48,20 @@ class LoadImageBase64:
         return (img, mask)
 
 
-class LoadMaskBase64:
+class LoadMaskBase64(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(s):
+    def define_schema(cls):
-        return {"required": {"mask": ("STRING", {"multiline": False})}}
+        return io.Schema(
+            node_id="ETN_LoadMaskBase64",
+            display_name="Load Mask (Base64)",
+            category="external_tooling",
+            inputs=[io.String.Input("mask", multiline=False)],
+            outputs=[io.Mask.Output(display_name="mask")],
+        )
 
-    RETURN_TYPES = ("MASK",)
+    @classmethod
-    CATEGORY = "external_tooling"
+    def execute(cls, mask: str):  # type: ignore
-    FUNCTION = "load_mask"
+        _strip_prefix(mask, "data:image/png;base64,")
-
-    def load_mask(self, mask):
         imgdata = base64.b64decode(mask)
         img = Image.open(BytesIO(imgdata))
         img = np.array(img).astype(np.float32) / 255.0
@@ -50,23 +69,24 @@ class LoadMaskBase64:
         if img.dim() == 3:  # RGB(A) input, use red channel
             img = img[:, :, 0]
         return (img.unsqueeze(0),)
-    
+
-class SendImageWebSocket:
+
+class SendImageWebSocket(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(s):
+    def define_schema(cls):
-        return {
+        return io.Schema(
-            "required": {
+            node_id="ETN_SendImageWebSocket",
-                "images": ("IMAGE",),
+            display_name="Send Image (WebSocket)",
-                "format": (["PNG", "JPEG"], {"default": "PNG"}),
+            category="external_tooling",
-            }
+            inputs=[
-        }
+                io.Image.Input("images"),
+                io.Combo.Input("format", options=["PNG", "JPEG"], default="PNG"),
+            ],
+            is_output_node=True,
+        )
 
-    RETURN_TYPES = ()
+    @classmethod
-    FUNCTION = "send_images"
+    def execute(cls, images: torch.Tensor, format: str):  # type: ignore
-    OUTPUT_NODE = True
-    CATEGORY = "external_tooling"
-
-    def send_images(self, images, format):
         results = []
         for tensor in images:
             array = 255.0 * tensor.cpu().numpy()
@@ -78,46 +98,161 @@ class SendImageWebSocket:
                 [format, image, None],
                 server.client_id,
             )
-            results.append(
+            results.append({
-                {"source": "websocket", "content-type": f"image/{format.lower()}", "type": "output"}
+                "source": "websocket",
-            )
+                "content-type": f"image/{format.lower()}",
+                "type": "output",
+            })
 
-        return {"ui": {"images": results}}
+        return io.NodeOutput(ui={"images": results})
 
-class CropImage:
+
-    """Deprecated, ComfyUI has an ImageCrop node now which does the same."""
+class ImageCache:
+    timeout = 600  # 10 minutes
+    max_size = 100 * 1024 * 1024  # 100 MB
+
+    @dataclass
+    class Entry:
+        data: bytes
+        content_type: str
+        timestamp: float
+        retrieved: int
+
+    class OldEntry(NamedTuple):
+        last_used: float
+        deleted: float
+        size: int
+        retrieved: int
+
+    def __init__(self):
+        self.images: dict[str, ImageCache.Entry] = {}
+        self.old: dict[str, ImageCache.OldEntry] = {}
+
+    def add(self, image: Image.Image, format: str):
+        key = uuid4().hex
+        with BytesIO() as output:
+            image.save(output, format=format, quality=95, compress_level=1)
+            image_data = output.getvalue()
+
+        self.insert(key, image_data, f"image/{format.lower()}")
+        return key
+
+    def insert(self, key: str, data: bytes, content_type: str):
+        self.images[key] = ImageCache.Entry(
+            data=data,
+            content_type=content_type,
+            timestamp=time.time(),
+            retrieved=0,
+        )
+
+    def get(self, key: str, extend: bool = False):
+        entry = self.images.get(key)
+        if entry is None:
+            if old := self.old.get(key):
+                now = time.time()
+                print(
+                    f"[comfyui-tooling-nodes] requested image {key} has been deleted ",
+                    f"(last used {now - old.last_used:.0f}s ago, deleted {now - old.deleted:.0f}s ago, "
+                    f"size {old.size / 1024**2:.1f}MB, retrieved {old.retrieved} times)",
+                )
+            return None, None
+        entry.retrieved += 1
+        if extend:
+            entry.timestamp = time.time()
+        self.prune()
+        return entry.data, entry.content_type
+
+    def prune(self):
+        total_size = sum(len(entry.data) for entry in self.images.values())
+        if total_size <= self.max_size:
+            return
+        # Remove least recently used entries until under max size
+        sorted_entries = sorted(self.images.items(), key=lambda item: item[1].timestamp)
+        now = time.time()
+        for key, entry in sorted_entries:
+            age = now - entry.timestamp
+            if age > self.timeout or (age > 60 and entry.retrieved > 0):
+                self.old[key] = ImageCache.OldEntry(
+                    entry.timestamp, now, len(entry.data), entry.retrieved
+                )
+                del self.images[key]
+                total_size -= len(entry.data)
+                if total_size <= self.max_size:
+                    break
+
+    def __contains__(self, key: str):
+        return key in self.images
+
+
+image_cache = ImageCache()
+
+
+class LoadImageCache(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_LoadImageCache",
+            display_name="Load Image from Cache",
+            category="external_tooling",
+            inputs=[io.String.Input("id", multiline=False)],
+            outputs=[io.Image.Output(display_name="image"), io.Mask.Output(display_name="mask")],
+        )
 
     @classmethod
-    def INPUT_TYPES(cls):
+    def execute(cls, id: str):  # type: ignore
-        return {
+        image_data, content_type = image_cache.get(id, extend=True)
-            "required": {
+        if image_data is None:
-                "image": ("IMAGE",),
+            raise ValueError(f"Image with ID {id} not found in cache.")
-                "x": (
-                    "INT",
-                    {"default": 0, "min": 0, "max": 8192, "step": 1},
-                ),
-                "y": (
-                    "INT",
-                    {"default": 0, "min": 0, "max": 8192, "step": 1},
-                ),
-                "width": (
-                    "INT",
-                    {"default": 512, "min": 1, "max": 8192, "step": 1},
-                ),
-                "height": (
-                    "INT",
-                    {"default": 512, "min": 1, "max": 8192, "step": 1},
-                ),
-            }
-        }
 
-    CATEGORY = "external_tooling"
+        img = Image.open(BytesIO(image_data))
-    RETURN_TYPES = ("IMAGE",)
-    FUNCTION = "crop"
 
-    def crop(self, image, x, y, width, height):
+        w, h = img.size
-        out = image[:, y : y + height, x : x + width, :]
+        c = len(img.getbands())
-        return (out,)
+        normalized = np.array(img).astype(np.float32) / 255.0
+        tensor = torch.from_numpy(normalized).reshape(1, h, w, c)
+        match c:
+            case 1:
+                image = tensor.expand(1, h, w, 3)
+                mask = tensor.reshape(1, h, w)
+            case 3:
+                image = tensor
+                mask = tensor[..., 0]
+            case 4:
+                image = tensor[..., :3]
+                mask = tensor[..., 3]
+
+        return io.NodeOutput(image, mask)
+
+
+class SaveImageCache(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_SaveImageCache",
+            display_name="Save Image to Cache",
+            category="external_tooling",
+            inputs=[
+                io.Image.Input("images"),
+                io.Combo.Input("format", options=["PNG", "JPEG"], default="PNG"),
+            ],
+            is_output_node=True,
+        )
+
+    @classmethod
+    def execute(cls, images: torch.Tensor, format: str):  # type: ignore
+        results = []
+        for tensor in images:
+            array = 255.0 * tensor.cpu().numpy()
+            image = Image.fromarray(np.clip(array, 0, 255).astype(np.uint8))
+            key = image_cache.add(image, format)
+
+            results.append({
+                "source": "http",
+                "id": key,
+                "content-type": f"image/{format.lower()}",
+                "type": "output",
+            })
+        return io.NodeOutput(ui={"images": results})
 
 
 def to_bchw(image: torch.Tensor):
@@ -136,21 +271,22 @@ def mask_batch(mask: torch.Tensor):
     return mask
 
 
-class ApplyMaskToImage:
+class ApplyMaskToImage(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {
+        return io.Schema(
-            "required": {
+            node_id="ETN_ApplyMaskToImage",
-                "image": ("IMAGE",),
+            display_name="Apply Mask to Image",
-                "mask": ("MASK",),
+            category="external_tooling",
-            }
+            inputs=[
-        }
+                io.Image.Input("image"),
+                io.Mask.Input("mask"),
+            ],
+            outputs=[io.Image.Output(display_name="masked")],
+        )
 
-    CATEGORY = "external_tooling"
+    @classmethod
-    RETURN_TYPES = ("IMAGE",)
+    def execute(cls, image: torch.Tensor, mask: torch.Tensor):  # type: ignore
-    FUNCTION = "apply_mask"
-
-    def apply_mask(self, image: torch.Tensor, mask: torch.Tensor):
         out = to_bchw(image)
         if out.shape[1] == 3:  # Assuming RGB images
             out = torch.cat([out, torch.ones_like(out[:, :1, :, :])], dim=1)
@@ -158,14 +294,138 @@ class ApplyMaskToImage:
 
         assert mask.ndim == 3, f"Mask should have shape [B, H, W]. {mask.shape}"
         assert out.ndim == 4, f"Image should have shape [B, C, H, W]. {out.shape}"
-        assert (
+        assert out.shape[-2:] == mask.shape[-2:], (
-            out.shape[-2:] == mask.shape[-2:]
+            f"Image size {out.shape[-2:]} must match mask size {mask.shape[-2:]}"
-        ), f"Image size {out.shape[-2:]} must match mask size {mask.shape[-2:]}"
+        )
         is_mask_batch = mask.shape[0] == out.shape[0]
 
         # Apply each mask in the batch to its corresponding image's alpha channel
         for i in range(out.shape[0]):
             alpha = mask[i] if is_mask_batch else mask[0]
-            out[i, 3, :, :] = alpha
+            out[i, 3, :, :] *= alpha
 
         return (to_bhwc(out),)
+
+
+class _ReferenceImageData(NamedTuple):
+    image: torch.Tensor
+    weight: float
+    range: tuple[float, float]
+
+
+class ReferenceImage(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_ReferenceImage",
+            display_name="Reference Image",
+            category="external_tooling",
+            inputs=[
+                io.Image.Input("image"),
+                io.Float.Input("weight", default=1.0, min=0.0, max=10.0),
+                io.Float.Input("range_start", default=0.0, min=0.0, max=1.0),
+                io.Float.Input("range_end", default=1.0, min=0.0, max=1.0),
+                io.Custom("ReferenceImage").Input("reference_images", optional=True),
+            ],
+            outputs=[io.Custom("ReferenceImage").Output(display_name="reference_images")],
+        )
+
+    @classmethod
+    def execute(  # type: ignore
+        cls,
+        image: torch.Tensor,
+        weight: float,
+        range_start: float,
+        range_end: float,
+        reference_images: list[_ReferenceImageData] | None = None,
+    ):
+        imgs = copy(reference_images) if reference_images is not None else []
+        imgs.append(_ReferenceImageData(image, weight, (range_start, range_end)))
+        return (imgs,)
+
+
+class ApplyReferenceImages(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_ApplyReferenceImages",
+            display_name="Apply Reference Images",
+            category="external_tooling",
+            inputs=[
+                io.Conditioning.Input("conditioning"),
+                io.ClipVision.Input("clip_vision"),
+                io.StyleModel.Input("style_model"),
+                io.Custom("ReferenceImage").Input("references"),
+            ],
+            outputs=[io.Conditioning.Output(display_name="conditioning")],
+        )
+
+    @classmethod
+    def execute(  # type: ignore
+        cls,
+        conditioning: list[list],
+        clip_vision: ClipVisionModel,
+        style_model: StyleModel,
+        references: list[_ReferenceImageData],
+    ):
+        delimiters = {0.0, 1.0}
+        delimiters |= set(r.range[0] for r in references)
+        delimiters |= set(r.range[1] for r in references)
+        delimiters = sorted(delimiters)
+        ranges = [(delimiters[i], delimiters[i + 1]) for i in range(len(delimiters) - 1)]
+
+        embeds = [_encode_image(r.image, clip_vision, style_model, r.weight) for r in references]
+        base = conditioning[0][0]
+        result = []
+        for start, end in ranges:
+            e = [
+                embeds[i]
+                for i, r in enumerate(references)
+                if r.range[0] <= start and r.range[1] >= end
+            ]
+            options = conditioning[0][1].copy()
+            options["start_percent"] = start
+            options["end_percent"] = end
+            result.append((torch.cat([base] + e, dim=1), options))
+
+        return (result,)
+
+
+def _encode_image(
+    image: torch.Tensor, clip_vision: ClipVisionModel, style_model: StyleModel, weight: float
+):
+    e = clip_vision.encode_image(image)
+    e = style_model.get_cond(e).flatten(start_dim=0, end_dim=1).unsqueeze(dim=0)
+    e = _downsample_image_cond(e, weight)
+    return e
+
+
+def _downsample_image_cond(cond: torch.Tensor, weight: float):
+    if weight >= 1.0:
+        return cond
+    elif weight <= 0.0:
+        return torch.zeros_like(cond)
+    elif weight >= 0.6:
+        factor = 2
+    elif weight >= 0.3:
+        factor = 3
+    else:
+        factor = 4
+
+    # Downsample the clip vision embedding to make it smaller, resulting in less impact
+    # compared to other conditioning.
+    # See https://github.com/kaibioinfo/ComfyUI_AdvancedRefluxControl
+    (b, t, h) = cond.shape
+    m = int(np.sqrt(t))
+    cond = F.interpolate(
+        cond.view(b, m, m, h).transpose(1, -1),
+        size=(m // factor, m // factor),
+        mode="area",
+    )
+    return cond.transpose(1, -1).reshape(b, -1, h)
+
+
+def _strip_prefix(s: str, prefix: str) -> str:
+    if s.startswith(prefix):
+        return s[len(prefix) :]
+    return s

nsfw.py

@@ -1,5 +1,4 @@
 from __future__ import annotations
-from weakref import ref as WeakRef
 from pathlib import Path
 from tqdm import tqdm
 import torch
@@ -8,6 +7,7 @@ import torch.nn.functional as F
 from torch import Tensor
 from transformers import CLIPImageProcessor, CLIPConfig, CLIPVisionModel, PreTrainedModel
 from kornia.filters import box_blur
+from comfy_api.latest import io
 
 from .nodes import to_bchw, to_bhwc
 
@@ -38,6 +38,10 @@ class CLIPSafetyChecker(PreTrainedModel):
         self.concept_embeds_weights = nn.Parameter(torch.ones(17), requires_grad=False)
         self.special_care_embeds_weights = nn.Parameter(torch.ones(3), requires_grad=False)
 
+        # Model requires post_init after transformers v4.57.3
+        if hasattr(self, "post_init"):
+            self.post_init()
+
     def forward(self, clip_input, images: Tensor, sensitivity: float):
         with torch.no_grad():
             image_batch = self.vision_model(clip_input)[1]
@@ -76,7 +80,7 @@ class CLIPSafetyChecker(PreTrainedModel):
 
 
 class CachedModels:
-    _instance: WeakRef | None = None
+    _instance: CachedModels | None = None
 
     def __init__(self):
         model_dir = Path(__file__).parent / "safetychecker"
@@ -91,11 +95,9 @@ class CachedModels:
 
     @classmethod
     def load(cls):
-        models = cls._instance and cls._instance()
+        if cls._instance is None:
-        if models is None:
+            cls._instance = CachedModels()
-            models = cls()
+        return cls._instance
-            cls._instance = WeakRef(models)
-        return models
 
     def download(self, url: str, target: Path):
         import requests
@@ -118,29 +120,26 @@ class CachedModels:
             ) from e
 
 
-class NSFWFilter:
+class NSFWFilter(io.ComfyNode):
-    models: CachedModels
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_NSFWFilter",
+            display_name="NSFW Filter",
+            category="external_tooling",
+            inputs=[
+                io.Image.Input("image"),
+                io.Float.Input("sensitivity", default=0.5, min=0.0, max=1.0, step=0.1),
+            ],
+            outputs=[io.Image.Output(display_name="image")],
+        )
 
     @classmethod
-    def INPUT_TYPES(cls):
+    def execute(cls, image: Tensor, sensitivity: float):
-        return {
+        models = CachedModels.load()
-            "required": {
-                "image": ("IMAGE",),
-                "sensitivity": ("FLOAT", {"default": 0.5, "min": 0.0, "max": 1.0, "step": 0.10}),
-            },
-        }
-
-    RETURN_TYPES = ("IMAGE",)
-    FUNCTION = "check"
-    CATEGORY = "external_tooling"
-
-    def __init__(self):
-        self.models = CachedModels.load()
-
-    def check(self, image, sensitivity):
         image = to_bchw(image)
-        input = self.models.feature_extractor(image, do_rescale=False, return_tensors="pt")
+        input = models.feature_extractor(image, do_rescale=False, return_tensors="pt")
-        filtered = self.models.safety_checker(
+        filtered = models.safety_checker(
             images=image, clip_input=input.pixel_values, sensitivity=sensitivity
         )
-        return (to_bhwc(filtered),)
+        return io.NodeOutput(to_bhwc(filtered))

pyproject.toml

@@ -1,12 +1,20 @@
 [project]
 name = "comfyui-tooling-nodes"
 description = "Provides nodes and server API extensions geared towards using ComfyUI as a backend for external tools."
-version = "1.5.0"
+version = "3.4.0"
 license = { file = "LICENSE" }
 
 [project.urls]
 Repository = "https://github.com/Acly/comfyui-tooling-nodes"
 
+[tool.ruff]
+target-version = "py311"
+line-length = 100
+preview = true
+
+[tool.ruff.lint]
+ignore = ["E741", "BLE001"]
+
 [tool.black]
 line-length = 100
 preview = true

region.py

@@ -1,13 +1,32 @@
+# Adapted from https://github.com/pamparamm/ComfyUI-ppm
 # Adapted from https://github.com/laksjdjf/cgem156-ComfyUI/blob/main/scripts/attention_couple/node.py
 # by @laksjdjf
 
 from __future__ import annotations
-from typing import NamedTuple
+from functools import partial
+from typing import Any, NamedTuple
 import torch
 import torch.nn.functional as F
 import math
 from torch import Tensor, Size
+import comfy.model_management
+import comfy.patcher_extension
 from comfy.model_patcher import ModelPatcher
+from comfy.model_base import Anima, CosmosPredict2
+from comfy.ldm.cosmos.predict2 import Attention as CosmosAttention
+from comfy.sampler_helpers import convert_cond
+from comfy.samplers import process_conds
+from comfy_api.latest import io
+
+
+COND = 0
+UNCOND = 1
+ANIMA_COUPLE_WRAPPER_KEY = "etn_attention_mask_anima"
+ANIMA_COUPLE_PATCH_KEY = "etn_attention_mask_patch"
+CONDS_COUPLE_KEY = "etn_couple_conds"
+COND_UNCOND_COUPLE_KEY = "etn_couple_cond_or_uncond"
+COUPLE_ACTIVE_KEY = "etn_couple_active"
+NUM_TOKENS_COUPLE_KEY = "etn_couple_num_tokens"
 
 
 def downsample_mask(mask: Tensor, batch: int, target_size: int, original_shape: Size) -> Tensor:
@@ -33,6 +52,12 @@ def downsample_mask(mask: Tensor, batch: int, target_size: int, original_shape:
     return result
 
 
+def reshape_mask(mask: Tensor, size: tuple[int, int], batch: int, target_size: int) -> Tensor:
+    result = F.interpolate(mask, size=size, mode="nearest")
+    result = result.view(mask.shape[0], target_size, 1)
+    return result.repeat_interleave(batch, dim=0)
+
+
 def lcm(a: int, b: int):
     return a * b // math.gcd(a, b)
 
@@ -65,105 +90,115 @@ class Region(NamedTuple):
         return result
 
 
-class BackgroundRegion:
+Regions = io.Custom("Regions")
+
+
+class BackgroundRegion(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {"required": {"conditioning": ("CONDITIONING",)}}
+        return io.Schema(
+            node_id="ETN_BackgroundRegion",
+            display_name="Background Region",
+            category="external_tooling/regions",
+            inputs=[io.Conditioning.Input("conditioning")],
+            outputs=[Regions.Output(display_name="regions")],
+        )
 
-    CATEGORY = "external_tooling/regions"
+    @classmethod
-    RETURN_TYPES = ("REGIONS",)
+    def execute(cls, conditioning: list):
-    FUNCTION = "define"
-
-    def define(self, conditioning: list):
         return (Region(None, None, conditioning),)
 
 
-class DefineRegion:
+class DefineRegion(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {
+        return io.Schema(
-            "required": {
+            node_id="ETN_DefineRegion",
-                "mask": ("MASK",),
+            display_name="Define Region",
-                "conditioning": ("CONDITIONING",),
+            category="external_tooling/regions",
-            },
+            inputs=[
-            "optional": {
+                io.Mask.Input("mask"),
-                "regions": ("REGIONS",),
+                io.Conditioning.Input("conditioning"),
-            },
+                Regions.Input("regions", optional=True),
-        }
+            ],
+            outputs=[Regions.Output(display_name="regions")],
+        )
 
-    CATEGORY = "external_tooling/regions"
+    @classmethod
-    RETURN_TYPES = ("REGIONS",)
+    def execute(cls, mask: Tensor, conditioning: list, regions: Region | None = None):
-    FUNCTION = "define"
-
-    def define(self, mask: Tensor, conditioning: list, regions: Region | None = None):
         if mask.dim() < 3:
             mask = mask.unsqueeze(0)
-        return (Region(regions, mask, conditioning),)
+        return io.NodeOutput(Region(regions, mask, conditioning))
 
 
-class ListRegionMasks:
+class ListRegionMasks(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {"required": {"regions": ("REGIONS",)}}
+        return io.Schema(
-
+            node_id="ETN_ListRegionMasks",
-    CATEGORY = "external_tooling/regions"
+            display_name="List Region Masks",
-    RETURN_TYPES = ("MASK",)
+            category="external_tooling/regions",
-    FUNCTION = "get_masks"
+            inputs=[Regions.Input("regions")],
-
+            outputs=[io.Mask.Output(display_name="masks")],
-    def get_masks(self, regions: Region):
+        )
-        return (torch.stack([r.mask for r in regions.preprocess()], dim=0),)
-
-
-class AttentionMask:
 
     @classmethod
-    def INPUT_TYPES(s):
+    def execute(cls, regions: Region):
-        return {
+        return io.NodeOutput(torch.stack([r.mask for r in regions.preprocess()], dim=0))
-            "required": {
-                "model": ("MODEL",),
-                "regions": ("REGIONS",),
-            }
-        }
 
-    RETURN_TYPES = ("MODEL",)
-    FUNCTION = "attention_mask"
-    CATEGORY = "external_tooling/regions"
 
-    mask: Tensor
+class AttentionMask(io.ComfyNode):
-    conds: list[Tensor]
+    @classmethod
-    batch_size: int
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_AttentionMask",
+            display_name="Regions Attention Mask",
+            category="external_tooling/regions",
+            inputs=[io.Model.Input("model"), Regions.Input("regions")],
+            outputs=[io.Model.Output(display_name="model")],
+        )
 
-    def attention_mask(self, model: ModelPatcher, regions: Region):
+    @classmethod
-        new_model = model.clone()
+    def execute(cls, model: ModelPatcher, regions: Region):
-        region_list = regions.preprocess()
+        return io.NodeOutput(AttentionMaskPatch.apply(model, regions))
-        num_conds = len(region_list)
 
+
+class AttentionMaskPatch:
+    def __init__(self, region_list: list[Region]):
         mask = torch.stack([r.mask for r in region_list], dim=0)
         mask_sum = mask.sum(dim=0, keepdim=True)
         assert mask_sum.sum() > 0, "There are areas that are zero in all masks."
         self.mask = mask / mask_sum
-
+        self.region_conds = [r.conditioning for r in region_list]
         self.conds = [r.conditioning[0][0] for r in region_list]
-        num_tokens = [cond.shape[1] for cond in self.conds]
+        self.num_tokens = [cond.shape[1] for cond in self.conds]
+        self.num_conds = len(region_list)
+        self.batch_size = 0
+
+    @staticmethod
+    def apply(model: ModelPatcher, regions: Region):
+        patch = AttentionMaskPatch(regions.preprocess())
+        if _is_anima_couple_model(model):
+            return patch.apply_anima(model)
 
         def attn2_patch(q: Tensor, k: Tensor, v: Tensor, extra_options: dict):
             assert k.mean() == v.mean(), "k and v must be the same."
             device, dtype = q.device, q.dtype
 
-            if self.conds[0].device != device or self.conds[0].dtype != dtype:
+            if patch.conds[0].device != device or patch.conds[0].dtype != dtype:
-                self.conds = [cond.to(device, dtype=dtype) for cond in self.conds]
+                patch.conds = [cond.to(device, dtype=dtype) for cond in patch.conds]
-            if self.mask.device != device or self.mask.dtype != dtype:
+            if patch.mask.device != device or patch.mask.dtype != dtype:
-                self.mask = self.mask.to(device, dtype=dtype)
+                patch.mask = patch.mask.to(device, dtype=dtype)
 
             cond_or_unconds = extra_options["cond_or_uncond"]
             num_chunks = len(cond_or_unconds)
-            self.batch_size = q.shape[0] // num_chunks
+            patch.batch_size = q.shape[0] // num_chunks
             q_chunks = q.chunk(num_chunks, dim=0)
             k_chunks = k.chunk(num_chunks, dim=0)
-            lcm_tokens = lcm_for_list(num_tokens + [k.shape[1]])
+            lcm_tokens = lcm_for_list(patch.num_tokens + [k.shape[1]])
             conds_tensor = [
-                cond.repeat(self.batch_size, lcm_tokens // num_tokens[i], 1)
+                cond.repeat(patch.batch_size, lcm_tokens // patch.num_tokens[i], 1)
-                for i, cond in enumerate(self.conds)
+                for i, cond in enumerate(patch.conds)
             ]
             conds_tensor = torch.cat(conds_tensor, dim=0)
 
@@ -174,9 +209,9 @@ class AttentionMask:
                     qs.insert(0, q_chunks[i])
                     ks.insert(0, k_target)
                 else:
-                    qs.insert(0, q_chunks[i].repeat(num_conds, 1, 1))
+                    qs.insert(0, q_chunks[i].repeat(patch.num_conds, 1, 1))
                     ks.insert(0, conds_tensor)
-                    for _ in range(num_conds - 1):
+                    for _ in range(patch.num_conds - 1):
                         cond_or_unconds.insert(i, 0)
 
             qs = torch.cat(qs, dim=0)
@@ -184,29 +219,218 @@ class AttentionMask:
             return qs, ks, ks
 
         def attn2_output_patch(out: Tensor, extra_options: dict):
+            num_conds = patch.num_conds
             cond_or_unconds = extra_options["cond_or_uncond"]
             mask_downsample = downsample_mask(
-                self.mask, self.batch_size, out.shape[1], extra_options["original_shape"]
+                patch.mask, patch.batch_size, out.shape[1], extra_options["original_shape"]
             )
             outputs: list[Tensor] = []
             pos = 0
             i = 0
             while i < len(cond_or_unconds):
                 if cond_or_unconds[i] == 1:  # uncond
-                    outputs.append(out[pos : pos + self.batch_size])
+                    outputs.append(out[pos : pos + patch.batch_size])
-                    pos += self.batch_size
+                    pos += patch.batch_size
                 else:
-                    masked = out[pos : pos + num_conds * self.batch_size] * mask_downsample
+                    masked = out[pos : pos + num_conds * patch.batch_size] * mask_downsample
-                    masked = masked.view(num_conds, self.batch_size, out.shape[1], out.shape[2])
+                    masked = masked.view(num_conds, patch.batch_size, out.shape[1], out.shape[2])
                     masked = masked.sum(dim=0)
                     outputs.append(masked)
-                    pos += num_conds * self.batch_size
+                    pos += num_conds * patch.batch_size
                     for _ in range(num_conds - 1):
                         cond_or_unconds.pop(i)
                 i += 1
 
             return torch.cat(outputs, dim=0)
 
+        new_model = model.clone()
         new_model.set_model_attn2_patch(attn2_patch)
         new_model.set_model_attn2_output_patch(attn2_output_patch)
-        return (new_model,)
+        new_model.set_attachments("etn_attention_mask", patch)
+        return new_model
+
+    def apply_anima(self, model: ModelPatcher):
+        new_model = model.clone()
+        _patch_cosmos_attention(new_model)
+
+        device = comfy.model_management.get_torch_device()
+        conds_converted = [convert_cond(cond)[0] for cond in self.region_conds]
+        new_model.add_wrapper_with_key(
+            comfy.patcher_extension.WrappersMP.SAMPLER_SAMPLE,
+            ANIMA_COUPLE_WRAPPER_KEY,
+            _anima_couple_sample_wrapper(conds_converted, device),
+        )
+        new_model.add_wrapper_with_key(
+            comfy.patcher_extension.WrappersMP.DIFFUSION_MODEL,
+            ANIMA_COUPLE_WRAPPER_KEY,
+            _anima_couple_diffusion_wrapper(self),
+        )
+        new_model.set_attachments("etn_attention_mask", self)
+        return new_model
+
+
+def _is_anima_couple_model(model: ModelPatcher) -> bool:
+    model_type = type(model.model)
+    return issubclass(model_type, (Anima, CosmosPredict2))
+
+
+def _anima_couple_sample_wrapper(conds_converted: list, device):
+    def sample_wrapper(executor, *args, **kwargs):
+        if len(conds_converted) > 0:
+            guider = args[0]
+            extra_options: dict[str, Any] = args[2]
+            seed: int = extra_options["seed"]
+            noise: Tensor = args[4]
+            latent_image: Tensor = args[5]
+            denoise_mask: Tensor | None = args[6]
+
+            conds_processed = process_conds(
+                guider.inner_model,
+                noise,
+                {"positive": conds_converted},
+                device,
+                latent_image,
+                denoise_mask,
+                seed,
+                latent_shapes=[latent_image.shape],
+            )["positive"]
+
+            conds_couple = [cond["model_conds"]["c_crossattn"].cond for cond in conds_processed]
+
+            model_options: dict[str, Any] = extra_options["model_options"]
+            transformer_options: dict[str, Any] = model_options.get("transformer_options", {}).copy()
+            transformer_options[CONDS_COUPLE_KEY] = conds_couple
+            transformer_options[NUM_TOKENS_COUPLE_KEY] = [cond.shape[1] for cond in conds_couple]
+            model_options["transformer_options"] = transformer_options
+
+        return executor(*args, **kwargs)
+
+    return sample_wrapper
+
+
+def _anima_couple_diffusion_wrapper(patch: AttentionMaskPatch):
+    def diffusion_wrapper(executor, *args, **kwargs):
+        anima_model = executor.class_obj
+        x: Tensor = args[0]
+        transformer_options: dict[str, Any] = kwargs.get("transformer_options", {}).copy()
+        patch_spatial = getattr(anima_model, "patch_spatial", 1)
+
+        activations_shape = list(x.shape)
+        activations_shape[-2] = activations_shape[-2] // patch_spatial
+        activations_shape[-1] = activations_shape[-1] // patch_spatial
+
+        transformer_options["activations_shape"] = activations_shape
+        transformer_options[ANIMA_COUPLE_PATCH_KEY] = patch
+        kwargs["transformer_options"] = transformer_options
+
+        return executor(*args, **kwargs)
+
+    return diffusion_wrapper
+
+
+def pre_cross_attention(
+    patch: AttentionMaskPatch,
+    transformer_options: dict,
+    x: Tensor,
+    context: Tensor,
+    rope_emb: Tensor | None,
+) -> tuple[Tensor, Tensor, Tensor | None, dict]:
+    transformer_options = transformer_options.copy()
+    if CONDS_COUPLE_KEY not in transformer_options:
+        transformer_options[COND_UNCOND_COUPLE_KEY] = list(transformer_options["cond_or_uncond"])
+        transformer_options[COUPLE_ACTIVE_KEY] = False
+        return x, context, rope_emb, transformer_options
+
+    conds: list[Tensor] = transformer_options[CONDS_COUPLE_KEY]
+    num_tokens_c: list[int] = transformer_options[NUM_TOKENS_COUPLE_KEY]
+    cond_or_uncond = transformer_options["cond_or_uncond"]
+
+    num_chunks = len(cond_or_uncond)
+    batch = x.shape[0] // num_chunks
+    x_chunks = x.chunk(num_chunks, dim=0)
+    c_chunks = context.chunk(num_chunks, dim=0)
+    lcm_tokens_c = lcm_for_list(num_tokens_c + [context.shape[1]])
+    conds_c_tensor = torch.cat(
+        [cond.repeat(batch, lcm_tokens_c // num_tokens_c[i], 1) for i, cond in enumerate(conds)],
+        dim=0,
+    )
+
+    xs, cs = [], []
+    cond_or_uncond_couple = []
+    for i, cond_type in enumerate(cond_or_uncond):
+        x_target = x_chunks[i]
+        c_target = c_chunks[i].repeat(1, lcm_tokens_c // context.shape[1], 1)
+        if cond_type == UNCOND:
+            xs.append(x_target)
+            cs.append(c_target)
+            cond_or_uncond_couple.append(UNCOND)
+        else:
+            xs.append(x_target.repeat(patch.num_conds, 1, 1))
+            cs.append(conds_c_tensor)
+            cond_or_uncond_couple.extend([COND] * patch.num_conds)
+
+    transformer_options[COND_UNCOND_COUPLE_KEY] = cond_or_uncond_couple
+    transformer_options[COUPLE_ACTIVE_KEY] = True
+
+    return torch.cat(xs, dim=0), torch.cat(cs, dim=0), rope_emb, transformer_options
+
+
+def cross_attention_output(patch: AttentionMaskPatch, transformer_options: dict, out: Tensor):
+    cond_or_uncond = transformer_options[COND_UNCOND_COUPLE_KEY]
+    size = tuple(transformer_options["activations_shape"][-2:])
+    batch = out.shape[0] // len(cond_or_uncond)
+    mask = patch.mask.to(out.device, dtype=out.dtype)
+    mask_downsample = reshape_mask(mask, size, batch, out.shape[1])
+
+    outputs = []
+    cond_outputs = []
+    i_cond = 0
+    for i, cond_type in enumerate(cond_or_uncond):
+        pos, next_pos = i * batch, (i + 1) * batch
+        if cond_type == UNCOND:
+            outputs.append(out[pos:next_pos])
+        else:
+            pos_cond, next_pos_cond = i_cond * batch, (i_cond + 1) * batch
+            cond_outputs.append(out[pos:next_pos] * mask_downsample[pos_cond:next_pos_cond])
+            i_cond += 1
+
+    if len(cond_outputs) > 0:
+        outputs.append(torch.stack(cond_outputs).sum(0))
+
+    return torch.cat(outputs, dim=0)
+
+
+def _patch_cosmos_attention(model_patcher: ModelPatcher):
+    cosmos_model = model_patcher.get_model_object("diffusion_model")
+    for block_name, block in (
+        (n, b)
+        for n, b in cosmos_model.named_modules()
+        if ("cross_attn" in n or "self_attn" in n) and isinstance(b, CosmosAttention)
+    ):
+        patch_name = f"diffusion_model.{block_name}.forward"
+        if patch_name not in model_patcher.object_patches:
+            model_patcher.add_object_patch(patch_name, partial(_cosmos_attention_forward_patched, block))
+
+
+def _cosmos_attention_forward_patched(
+    self,
+    x: Tensor,
+    context: Tensor | None = None,
+    rope_emb: Tensor | None = None,
+    transformer_options: dict | None = None,
+) -> Tensor:
+    transformer_options = transformer_options if transformer_options is not None else {}
+    patch: AttentionMaskPatch | None = transformer_options.get(ANIMA_COUPLE_PATCH_KEY)
+
+    if context is not None and patch is not None:
+        x, context, rope_emb, transformer_options = pre_cross_attention(
+            patch, transformer_options, x, context, rope_emb
+        )
+
+    q, k, v = self.compute_qkv(x, context, rope_emb=rope_emb)
+    output = self.compute_attention(q, k, v, transformer_options=transformer_options)
+
+    if context is not None and patch is not None and transformer_options.get(COUPLE_ACTIVE_KEY, False):
+        output = cross_attention_output(patch, transformer_options, output)
+
+    return output

tile.py

@@ -3,49 +3,29 @@ import numpy as np
 import numpy.typing as npt
 import torch
 from torch import Tensor
+from comfy_api.latest import io
 
 IntArray = npt.NDArray[np.int_]
 
 
 class TileLayout:
-    @classmethod
+    def __init__(
-    def INPUT_TYPES(cls):
+        self, image: Tensor, min_tile_size: int, padding: int, blending: int, multiple: int
-        return {
+    ):
-            "required": {
+        assert all([x % multiple == 0 for x in image.shape[-3:-1]]), (
-                "image": ("IMAGE",),
+            "Image size must be divisible by multiple"
-                "min_tile_size": ("INT", {"default": 512, "min": 64, "max": 8192, "step": 8}),
+        )
-                "padding": ("INT", {"default": 32, "min": 0, "max": 8192, "step": 8}),
+        assert min_tile_size % multiple == 0, "Tile size must be divisible by multiple"
-                "blending": ("INT", {"default": 8, "min": 0, "max": 256, "step": 8}),
+        assert blending <= padding, "Blending must be smaller than padding"
-            }
-        }
 
-    CATEGORY = "external_tooling/tiles"
+        self.image_size: IntArray = np.array(image.shape[-3:-1])
-    RETURN_TYPES = ("TILE_LAYOUT",)
+        self.padding: int = padding
-    FUNCTION = "node"
+        self.blending: int = blending
-
+        self.tile_count: IntArray = np.maximum(1, self.image_size // (min_tile_size - 2 * padding))
-    image_size: IntArray
-    tile_size: IntArray
-    padding: int
-    blending: int
-    tile_count: IntArray
-
-    def node(self, image: Tensor, min_tile_size: int, padding: int, blending: int):
-        self.init(image, min_tile_size, padding, blending)
-        return (self,)
-
-    def init(self, image: Tensor, min_tile_size: int, padding: int, blending: int):
-        assert all([x % 8 == 0 for x in image.shape[-3:-1]]), "Image size must be divisible by 8"
-        assert min_tile_size % 8 == 0, "Tile size must be divisible by 8"
-        assert blending < padding, "Blending must be smaller than padding"
-
-        self.image_size = np.array(image.shape[-3:-1])
-        self.padding = padding
-        self.blending = blending
-        self.tile_count = np.maximum(1, self.image_size // (min_tile_size - 2 * padding))
 
         image_size_with_overlap = self.image_size + (self.tile_count - 1) * 2 * padding
         tile_size = np.ceil(image_size_with_overlap / self.tile_count)
-        self.tile_size = (np.ceil(tile_size / 8) * 8).astype(int)
+        self.tile_size: IntArray = (np.ceil(tile_size / multiple) * multiple).astype(int)
 
     def size(self, coord: IntArray):
         return self.end(coord) - self.start(coord)
@@ -96,80 +76,109 @@ class TileLayout:
         image[rect] = (1 - mask) * image[rect] + mask * tile
 
 
-class ExtractImageTile:
+class CreateTileLayout(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {
+        return io.Schema(
-            "required": {
+            node_id="ETN_TileLayout",
-                "image": ("IMAGE",),
+            display_name="Create Tile Layout",
-                "layout": ("TILE_LAYOUT",),
+            category="external_tooling/tiles",
-                "index": ("INT", {"min": 0}),
+            inputs=[
-            }
+                io.Image.Input("image"),
-        }
+                io.Int.Input("min_tile_size", default=512, min=64, max=8192, step=8),
+                io.Int.Input("padding", default=32, min=0, max=8192, step=8),
+                io.Int.Input("blending", default=8, min=0, max=256, step=8),
+                io.Int.Input("multiple", default=8, min=1, max=1024, step=1),
+            ],
+            outputs=[io.Custom("TileLayout").Output(display_name="layout")],
+        )
 
-    CATEGORY = "external_tooling/tiles"
-    RETURN_TYPES = ("IMAGE",)
-    FUNCTION = "slice"
-
-    def slice(self, image: Tensor, layout: TileLayout, index: int):
-        return (layout.tile(image, index),)
-
-
-class ExtractMaskTile:
     @classmethod
-    def INPUT_TYPES(cls):
+    def execute(cls, image: Tensor, min_tile_size: int, padding: int, blending: int, multiple: int):
-        return {
+        return io.NodeOutput(TileLayout(image, min_tile_size, padding, blending, multiple))
-            "required": {
-                "mask": ("MASK",),
-                "layout": ("TILE_LAYOUT",),
-                "index": ("INT", {"min": 0}),
-            }
-        }
 
-    CATEGORY = "external_tooling/tiles"
-    RETURN_TYPES = ("MASK",)
-    FUNCTION = "slice"
 
-    def slice(self, mask: Tensor, layout: TileLayout, index: int):
+class ExtractImageTile(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_ExtractImageTile",
+            display_name="Extract Image Tile",
+            category="external_tooling/tiles",
+            inputs=[
+                io.Image.Input("image"),
+                io.Custom("TileLayout").Input("layout"),
+                io.Int.Input("index", default=0, min=0),
+            ],
+            outputs=[io.Image.Output(display_name="tile")],
+        )
+
+    @classmethod
+    def execute(cls, image: Tensor, layout: TileLayout, index: int):
+        return io.NodeOutput(layout.tile(image, index))
+
+
+class ExtractMaskTile(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_ExtractMaskTile",
+            display_name="Extract Mask Tile",
+            category="external_tooling/tiles",
+            inputs=[
+                io.Mask.Input("mask"),
+                io.Custom("TileLayout").Input("layout"),
+                io.Int.Input("index", default=0, min=0),
+            ],
+            outputs=[io.Mask.Output(display_name="tile")],
+        )
+
+    @classmethod
+    def execute(cls, mask: Tensor, layout: TileLayout, index: int):
         tile = layout.tile(mask.unsqueeze(3), index)
-        return (tile.squeeze(3),)
+        return io.NodeOutput(tile.squeeze(3))
 
 
-class GenerateTileMask:
+class GenerateTileMask(io.ComfyNode):
     @classmethod
-    def INPUT_TYPES(cls):
+    def define_schema(cls):
-        return {
+        return io.Schema(
-            "required": {"layout": ("TILE_LAYOUT",), "index": ("INT", {"min": 0})},
+            node_id="ETN_GenerateTileMask",
-            "optional": {"blend": ("BOOLEAN",)},
+            display_name="Generate Tile Mask",
-        }
+            category="external_tooling/tiles",
+            inputs=[
+                io.Custom("TileLayout").Input("layout"),
+                io.Int.Input("index", default=0, min=0),
+                io.Boolean.Input("blend", default=False, optional=True),
+            ],
+            outputs=[io.Mask.Output(display_name="mask")],
+        )
 
-    CATEGORY = "external_tooling/tiles"
-    RETURN_TYPES = ("MASK",)
-    FUNCTION = "generate"
-
-    def generate(self, layout: TileLayout, index: int, blend: bool = False):
-        return (layout.mask(layout.coord(index), blend=blend),)
-
-
-class MergeImageTile:
     @classmethod
-    def INPUT_TYPES(cls):
+    def execute(cls, layout: TileLayout, index: int, blend: bool = False):
-        return {
+        return io.NodeOutput(layout.mask(layout.coord(index), blend=blend))
-            "required": {
-                "image": ("IMAGE",),
-                "layout": ("TILE_LAYOUT",),
-                "index": ("INT", {"min": 0}),
-                "tile": ("IMAGE",),
-            }
-        }
 
-    CATEGORY = "external_tooling/tiles"
-    RETURN_TYPES = ("IMAGE",)
-    FUNCTION = "merge"
 
-    def merge(self, image: Tensor, layout: TileLayout, index: int, tile: Tensor):
+class MergeImageTile(io.ComfyNode):
+    @classmethod
+    def define_schema(cls):
+        return io.Schema(
+            node_id="ETN_MergeImageTile",
+            display_name="Merge Image Tile",
+            category="external_tooling/tiles",
+            inputs=[
+                io.Image.Input("image"),
+                io.Custom("TileLayout").Input("layout"),
+                io.Int.Input("index", default=0, min=0),
+                io.Image.Input("tile"),
+            ],
+            outputs=[io.Image.Output(display_name="image")],
+        )
+
+    @classmethod
+    def execute(cls, image: Tensor, layout: TileLayout, index: int, tile: Tensor):
         assert index < layout.total_count, f"Index {index} out of range"
         if index == 0:
             image = image.clone()
         layout.merge(image, index, tile)
-        return (image,)
+        return io.NodeOutput(image)

translation.py

@@ -10,6 +10,7 @@ from __future__ import annotations
 import re
 from functools import cache
 from typing import NamedTuple
+from comfy_api.latest import io
 
 
 @cache
@@ -43,7 +44,7 @@ def translate_chunk(text: str, language: str):
                 (p for p in available if p.from_code == language and p.to_code == target), None
             )
             assert pkg, f"Couldn't find package for translation from {language}"
-            print("Downloading and installing translation package", pkg)
+            # print("Downloading and installing translation package", pkg)  # this will cause encoding errors
             pkg.install()
 
         text, embeddings = _extract_embeddings(text)
@@ -61,17 +62,20 @@ def translate(text: str):
     return " ".join(translate_chunk(c.text, c.lang) for c in chunks)
 
 
-class Translate:
+class Translate(io.ComfyNode):
-    @staticmethod
+    @classmethod
-    def INPUT_TYPES():
+    def define_schema(cls):
-        return {"required": {"text": ("STRING", {"multiline": True})}}
+        return io.Schema(
+            node_id="ETN_Translate",
+            display_name="Translate Text",
+            category="external_tooling",
+            inputs=[io.String.Input("text", multiline=True)],
+            outputs=[io.String.Output(display_name="translation")],
+        )
 
-    CATEGORY = "external_tooling"
+    @classmethod
-    RETURN_TYPES = ("STRING",)
+    def execute(cls, text: str):
-    FUNCTION = "translate"
+        return io.NodeOutput(translate(text))
-
-    def translate(self, text: str):
-        return (translate(text),)
 
 
 _lang_regex = re.compile(r"(lang:\w\w)")