UNPKG

comfyui-mcp

Version:

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),

63 lines (55 loc) • 3.96 kB
# LTX-2.3 3-stage distill install manifest — consumable by apply_manifest (src/services/manifest.ts). # Ground truth: jcd315/comfyui-mcp-muse workflows/ltx23_distill_3stage.json # (commits 63f3b2a feat, 6d2ccd2 rename, d087cdf resolution-defaults fix). # Every model below is referenced by a loader node in workflow.json; URLs verified # via HEAD request 2026-07-12. Sources: Lightricks (official), Kijai/LTX2.3_comfy # (transformer-only fp8 split + taeltx preview VAE), Comfy-Org/ltx-2 (gemma + # heretic LoRA), Aitrepreneur/FLX mirror (split VAEs + text projection, same # files the sibling ltx-2.3 packs pull). pip: - imageio-ffmpeg # VHS_VideoCombine needs an ffmpeg for the mux # Same node set as the sibling ltx-2.3 packs (the workflow derives from the same # ULTRA ecosystem): LTXVideo (LTXV*/LTX2_NAG/SageAttention patch/ManualSigmas), # KJNodes (ImageResizeKJv2/SimpleCalculatorKJ/VAELoaderKJ/INTConstant), rgthree # (Power Lora Loader), VHS (LoadImagePath/VideoCombine), ComfyMath, plus the rest # of the shared ULTRA toolchain (ResizeImagesByLongerEdge et al.). custom_nodes: - https://github.com/city96/ComfyUI-GGUF - https://github.com/rgthree/rgthree-comfy - https://github.com/yolain/ComfyUI-Easy-Use - https://github.com/kijai/ComfyUI-KJNodes - https://github.com/ClownsharkBatwing/RES4LYF - https://github.com/Lightricks/ComfyUI-LTXVideo - https://github.com/pythongosssss/ComfyUI-Custom-Scripts - https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite - https://github.com/kijai/ComfyUI-WanVideoWrapper - https://github.com/ltdrdata/ComfyUI-Impact-Pack - https://github.com/TTPlanetPig/Comfyui_TTP_Toolset - https://github.com/evanspearman/ComfyMath - https://github.com/WhatDreamsCost/WhatDreamsCost-ComfyUI models: # --- UNet: 22B dev transformer-only, fp8 input-scaled (~23 GB) --- - url: https://huggingface.co/Kijai/LTX2.3_comfy/resolve/main/diffusion_models/ltx-2-3-22b-dev_transformer_only_fp8_input_scaled.safetensors local_path: diffusion_models/ltx-2-3-22b-dev_transformer_only_fp8_input_scaled.safetensors # --- Text encoder: Gemma-3-12B (bf16) + LTX text projection --- - url: https://huggingface.co/Comfy-Org/ltx-2/resolve/main/split_files/text_encoders/gemma_3_12B_it.safetensors local_path: text_encoders/gemma_3_12B_it.safetensors - url: https://huggingface.co/Aitrepreneur/FLX/resolve/main/ltx-2.3_text_projection_bf16.safetensors?download=true local_path: text_encoders/ltx-2.3_text_projection_bf16.safetensors # --- VAE (separate video + audio, plus taeltx fast preview decoder) --- - url: https://huggingface.co/Aitrepreneur/FLX/resolve/main/LTX23_video_vae_bf16.safetensors?download=true local_path: vae/LTX23_video_vae_bf16.safetensors - url: https://huggingface.co/Aitrepreneur/FLX/resolve/main/LTX23_audio_vae_bf16.safetensors?download=true local_path: vae/LTX23_audio_vae_bf16.safetensors - url: https://huggingface.co/Kijai/LTX2.3_comfy/resolve/main/vae/taeltx2_3.safetensors local_path: vae/taeltx2_3.safetensors # --- Spatial upscalers (pass 2: x2, pass 3: x1.5) --- - url: https://huggingface.co/Lightricks/LTX-2.3/resolve/main/ltx-2.3-spatial-upscaler-x2-1.1.safetensors local_path: latent_upscale_models/ltx-2.3-spatial-upscaler-x2-1.1.safetensors - url: https://huggingface.co/Lightricks/LTX-2.3/resolve/main/ltx-2.3-spatial-upscaler-x1.5-1.0.safetensors local_path: latent_upscale_models/ltx-2.3-spatial-upscaler-x1.5-1.0.safetensors # --- LoRAs: distilled speed LoRA @ 0.6 + gemma abliterated-heretic @ 1.0 --- - url: https://huggingface.co/Lightricks/LTX-2.3/resolve/main/ltx-2.3-22b-distilled-lora-384.safetensors local_path: loras/ltx-2.3-22b-distilled-lora-384.safetensors - url: https://huggingface.co/Comfy-Org/ltx-2/resolve/main/split_files/loras/gemma-3-12b-it-abliterated_heretic_lora_rank64_bf16.safetensors local_path: loras/gemma-3-12b-it-abliterated_heretic_lora_rank64_bf16.safetensors