comfyui-mcp 0.52.175 → 0.52.176

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (7) hide show
  1. package/package.json +1 -1
  2. package/packs/wan-multitalk/install-runpod.sh +2 -1
  3. package/packs/wan-multitalk/install-windows.bat +2 -1
  4. package/packs/wan-multitalk/manifest.yaml +9 -6
  5. package/packs/wan-multitalk/pack.yaml +4 -1
  6. package/packs/wan-multitalk/workflow.json +2 -2
  7. package/plugin/skills/wan-multitalk/SKILL.md +9 -7
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "comfyui-mcp",
3
- "version": "0.52.175",
3
+ "version": "0.52.176",
4
4
  "mcpName": "io.github.artokun/comfyui-mcp",
5
5
  "description": "Local-first, agent-native control plane for ComfyUI — MCP server + autonomous sidebar agent that drives your live graph in natural language on ANY LLM: Claude/ChatGPT/Gemini on your subscription (no API key), free local models via Ollama (fully offline), or any hosted model via an OpenAI-compatible endpoint (DeepSeek, GLM, MiMo, OpenRouter). Generate images, video & audio, author and run workflows, manage models and custom nodes; a compact tool-router mode keeps even 4B models effective, and a built-in LLM Arena benchmarks them on real ComfyUI tasks. Also a Claude Code plugin with skills, slash commands, and installer packs. Local, LAN, VPS, or Comfy Cloud.",
6
6
  "homepage": "https://comfyui-mcp.artokun.io/docs",
package/packs/wan-multitalk/install-runpod.sh CHANGED
@@ -28,6 +28,7 @@ echo "-------- custom nodes --------"
28
28
  clone "ComfyUI-WanVideoWrapper" "https://github.com/kijai/ComfyUI-WanVideoWrapper.git"
29
29
  clone "ComfyUI-KJNodes" "https://github.com/kijai/ComfyUI-KJNodes.git"
30
30
  clone "ComfyUI-VideoHelperSuite" "https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite.git"
31
+ clone "audio-separation-nodes-comfyui" "https://github.com/christian-byrne/audio-separation-nodes-comfyui.git"
31
32
 
32
33
  echo "-------- pip (manifest extras) --------"
33
34
  "$PY" -m pip install "imageio-ffmpeg"
@@ -35,7 +36,7 @@ echo "-------- pip (manifest extras) --------"
35
36
  echo "-------- models --------"
36
37
  grab "models/diffusion_models/Wan2.1_14b_Image_to_Video_480p_GGUF_Q8.gguf" "https://huggingface.co/city96/Wan2.1-I2V-14B-480P-gguf/resolve/main/wan2.1-i2v-14b-480p-Q8_0.gguf"
37
38
  grab "models/diffusion_models/WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors" "https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors"
38
- grab "models/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors" "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors"
39
+ grab "models/text_encoders/umt5_xxl_fp16.safetensors" "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors"
39
40
  grab "models/vae/Wan2_1_VAE_bf16.safetensors" "https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors"
40
41
  grab "models/clip_vision/clip_vision_h.safetensors" "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors"
41
42
  grab "models/loras/Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors" "https://huggingface.co/lightx2v/Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v/resolve/main/loras/Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors"
package/packs/wan-multitalk/install-windows.bat CHANGED
@@ -18,6 +18,7 @@ echo -------- custom nodes --------
18
18
  call :clone "ComfyUI-WanVideoWrapper" "https://github.com/kijai/ComfyUI-WanVideoWrapper.git"
19
19
  call :clone "ComfyUI-KJNodes" "https://github.com/kijai/ComfyUI-KJNodes.git"
20
20
  call :clone "ComfyUI-VideoHelperSuite" "https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite.git"
21
+ call :clone "audio-separation-nodes-comfyui" "https://github.com/christian-byrne/audio-separation-nodes-comfyui.git"
21
22
 
22
23
  echo -------- pip (manifest extras) --------
23
24
  "%PY%" -m pip install "imageio-ffmpeg"
@@ -25,7 +26,7 @@ echo -------- pip (manifest extras) --------
25
26
  echo -------- models --------
26
27
  call :grab "models\diffusion_models\Wan2.1_14b_Image_to_Video_480p_GGUF_Q8.gguf" "https://huggingface.co/city96/Wan2.1-I2V-14B-480P-gguf/resolve/main/wan2.1-i2v-14b-480p-Q8_0.gguf"
27
28
  call :grab "models\diffusion_models\WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors" "https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors"
28
- call :grab "models\text_encoders\umt5_xxl_fp8_e4m3fn_scaled.safetensors" "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors"
29
+ call :grab "models\text_encoders\umt5_xxl_fp16.safetensors" "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors"
29
30
  call :grab "models\vae\Wan2_1_VAE_bf16.safetensors" "https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors"
30
31
  call :grab "models\clip_vision\clip_vision_h.safetensors" "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/clip_vision/clip_vision_h.safetensors"
31
32
  call :grab "models\loras\Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors" "https://huggingface.co/lightx2v/Wan2.1-I2V-14B-480P-StepDistill-CfgDistill-Lightx2v/resolve/main/loras/Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors"
package/packs/wan-multitalk/manifest.yaml CHANGED
@@ -9,9 +9,11 @@
9
9
  # All 6 model URLs below are HEAD-verified live (packs:check-urls, 6/6 OK). Wav2Vec
10
10
  # is auto-downloaded by DownloadAndLoadWav2VecModel (no manifest entry needed).
11
11
  # The GGUF downloads under city96's name and is saved as the workflow's local name.
12
- # NOT yet render-verified — do a live render on a local ComfyUI before shipping.
13
- # If AudioSeparation/AudioCrop don't resolve from WanVideoWrapper, add the audio
14
- # node pack that provides them.
12
+ # The workflow's AudioSeparation/AudioCrop nodes come from the explicitly declared
13
+ # audio-separation-nodes-comfyui pack below; they are not WanVideoWrapper nodes.
14
+ # The bundled WanVideoWrapper revision rejects scaled_fp8 UMT5 checkpoints, so use
15
+ # the compatible FP16 pairing instead. NOT yet render-verified — do
16
+ # a live render on a local ComfyUI before shipping.
15
17
 
16
18
  pip:
17
19
  - imageio-ffmpeg # VHS_VideoCombine needs an ffmpeg for the mux
@@ -20,6 +22,7 @@ custom_nodes:
20
22
  - https://github.com/kijai/ComfyUI-WanVideoWrapper.git # WanVideo*/MultiTalk*/Wav2Vec/T5 nodes
21
23
  - https://github.com/kijai/ComfyUI-KJNodes.git # ImageResizeKJv2
22
24
  - https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite.git # VHS_VideoCombine
25
+ - https://github.com/christian-byrne/audio-separation-nodes-comfyui.git # AudioSeparation + AudioCrop
23
26
 
24
27
  models:
25
28
  # --- WAN 2.1 14B Image-to-Video 480p, GGUF Q8 (main transformer) ---
@@ -32,9 +35,9 @@ models:
32
35
  - url: https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors
33
36
  local_path: diffusion_models/WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors
34
37
 
35
- # --- umt5-xxl text encoder (fp8 scaled) ---
36
- - url: https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors
37
- local_path: text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors
38
+ # --- umt5-xxl text encoder (fp16; compatible with LoadWanVideoT5TextEncoder) ---
39
+ - url: https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/resolve/main/split_files/text_encoders/umt5_xxl_fp16.safetensors
40
+ local_path: text_encoders/umt5_xxl_fp16.safetensors
38
41
 
39
42
  # --- WAN 2.1 VAE (bf16) ---
40
43
  - url: https://huggingface.co/Kijai/WanVideo_comfy/resolve/main/Wan2_1_VAE_bf16.safetensors
package/packs/wan-multitalk/pack.yaml CHANGED
@@ -18,7 +18,10 @@ sources:
18
18
  notes:
19
19
  - Run the generated installer from your ComfyUI root.
20
20
  - Needs kijai's ComfyUI-WanVideoWrapper (all WanVideo*/MultiTalk* nodes), KJNodes
21
- (ImageResizeKJv2) and VideoHelperSuite (VHS_VideoCombine).
21
+ (ImageResizeKJv2), VideoHelperSuite (VHS_VideoCombine), and
22
+ christian-byrne's audio-separation-nodes-comfyui (AudioSeparation + AudioCrop).
23
+ - Uses umt5_xxl_fp16.safetensors because the bundled WanVideoWrapper loader
24
+ rejects the scaled_fp8 UMT5 checkpoint; do not substitute generic t5xxl weights.
22
25
  - Wav2Vec is auto-downloaded on first run by DownloadAndLoadWav2VecModel — no
23
26
  manual model needed for it.
24
27
  - VERIFY the model URLs against your setup before a clean install — the GGUF and
package/packs/wan-multitalk/workflow.json CHANGED
@@ -417,7 +417,7 @@
417
417
  "ver": "058286fc0f3b0651a2f6b68309df3f06e8332cc0"
418
418
  },
419
419
  "widgets_values": [
420
- "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
420
+ "umt5_xxl_fp16.safetensors",
421
421
  "bf16",
422
422
  "offload_device",
423
423
  "disabled"
@@ -1429,4 +1429,4 @@
1429
1429
  "VHS_KeepIntermediate": true
1430
1430
  },
1431
1431
  "version": 0.4
1432
- }
1432
+ }
package/plugin/skills/wan-multitalk/SKILL.md CHANGED
@@ -42,8 +42,8 @@ Key nodes (all kijai WanVideoWrapper unless noted):
42
42
  embeddings that steer the mouth and expression.
43
43
  - **MultiTalkModelLoader** + **WanVideoImageToVideoMultiTalk.** The MultiTalk head
44
44
  on top of the WAN I2V model.
45
- - **AudioSeparation.** Isolate the voice from music/noise before embedding (cleaner
46
- lip-sync). **AudioCrop.** Trim to the segment you want to animate.
45
+ - **AudioSeparation** and **AudioCrop** (audio-separation-nodes-comfyui). Isolate the
46
+ voice from music/noise before embedding and trim the segment you want to animate.
47
47
  - **ImageResizeKJv2** (KJNodes), **VHS_VideoCombine** (VideoHelperSuite). Resize and
48
48
  mux to mp4.
49
49
 
@@ -52,15 +52,17 @@ Key nodes (all kijai WanVideoWrapper unless noted):
52
52
  | File | Loader | Folder |
53
53
  |------|--------|--------|
54
54
  | `Wan2.1_14b_Image_to_Video_480p_GGUF_Q8.gguf` | WanVideoModelLoader | `diffusion_models/` |
55
- | `WanVideo_2_1_Multitalk_14B_fp32.safetensors` | MultiTalkModelLoader | `diffusion_models/` |
56
- | `umt5_xxl_fp8_e4m3fn_scaled.safetensors` | LoadWanVideoT5TextEncoder | `text_encoders/` |
55
+ | `WanVideo_2_1_Multitalk_14B_fp8_e4m3fn.safetensors` | MultiTalkModelLoader | `diffusion_models/` |
56
+ | `umt5_xxl_fp16.safetensors` | LoadWanVideoT5TextEncoder | `text_encoders/` |
57
57
  | `Wan2_1_VAE_bf16.safetensors` | WanVideoVAELoader | `vae/` |
58
58
  | `clip_vision_h.safetensors` | CLIPVisionLoader | `clip_vision/` |
59
- | `Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64_fixed.safetensors` | WanVideoLoraSelect | `loras/` |
59
+ | `Wan21_I2V_14B_lightx2v_cfg_step_distill_lora_rank64.safetensors` | WanVideoLoraSelect | `loras/` |
60
60
 
61
61
  Sources: kijai `Kijai/WanVideo_comfy`, MeiGen-AI `MeiGen-AI/MeiGen-MultiTalk`, GGUF
62
- `city96/Wan2.1-I2V-14B-480P-gguf`. See `packs/wan-multitalk/manifest.yaml` (some URLs
63
- are best-effort; verify per mirror). Wav2Vec auto-downloads.
62
+ `city96/Wan2.1-I2V-14B-480P-gguf`, and Comfy-Org's repackaged UMT5. See
63
+ `packs/wan-multitalk/manifest.yaml` (some URLs are best-effort; verify per mirror).
64
+ Wav2Vec auto-downloads. The bundled WanVideoWrapper loader rejects the scaled_fp8
65
+ UMT5 checkpoint; use the UMT5 fp16 file above, not generic `t5xxl_fp16` weights.
64
66
 
65
67
  ## Inputs & key parameters
66
68