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),
31 lines (30 loc) • 1.39 kB
YAML
name: anima-img2img
display_name: ANIMA 1.0 — Image to Image (ControlNet)
family: anima
kind: img2img
description: >-
ANIMA 1.0 (circlestone-labs, ~2B DiT anime/illustration base on NVIDIA
Cosmos-Predict2-2B) image-to-image with Anima-LLLite ControlNet guidance —
depth/openpose preprocessing steer the redraw from a source image. Single
pipeline sliced from the ANIMA monolith (no group toggles), with the Turbo
LoRA preset, latent upscaler, and Impact-Pack face/hand detailing built in.
vram: "<6GB"
workflow: workflow.json
skill: anima-base
launch_args: []
sources:
patreon: https://www.patreon.com/posts/157355820
model: https://huggingface.co/circlestone-labs/Anima
mirror: https://huggingface.co/Aitrepreneur/FLX
notes:
- Run the generated installer from your ComfyUI root (folder with custom_nodes/
and models/).
- Sliced + un-bypassed from the monolith's "IMAGE TO IMAGE TURBO CONTROLNET".
- Uses the kohya-ss ComfyUI-Anima-LLLite node (AnimaLLLiteApply_sdscripts) for ControlNet.
- Depth/openpose preprocessor weights (depth_anything_v2_vitl,
dw-ll_ucoco_384) are auto-downloaded by comfyui_controlnet_aux on first run.
- Turbo LoRA preset is the shipped default — 12 steps, CFG 1.0, er_sde/simple.
post_install:
- Restart ComfyUI (or ComfyUI-Manager -> Install missing custom nodes for any
node Python deps).
- Load workflow.json.