comfyui-mcp/packs/qwen-image/pack.yaml

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),

name: qwen-image
display_name: Qwen-Image (20B T2I + WAN combo)
family: qwen
kind: txt2img
description: >-
  Qwen-Image 20B text-to-image "Ultra/Combo" pack — a distilled Qwen-Image
  diffusion model (GGUF) with the Qwen2.5-VL 7B text encoder, paired with a
  WAN 2.2 T2V A14B low-noise refine stage for a two-model combo pipeline.
  Includes 8-step / 4-step Lightning acceleration LoRAs and the GGUF quant menu
  (Q4_K_S / Q5_K_S / Q8_0) so it scales from sub-12GB GPUs to 24GB+.
vram: "12GB+ (Q4_K_S); 24GB+ for Q8_0"
workflow: null
skill: qwen-txt2img
launch_args: []
sources:
  model: https://huggingface.co/Comfy-Org/Qwen-Image_ComfyUI
  mirror: https://huggingface.co/Aitrepreneur/FLX
notes:
  - >-
    Run the generated installer from your ComfyUI root (the folder containing
    custom_nodes/ and models/).
  - >-
    Workflow is NOT bundled. Load QWEN_COMBO_ULTRA_WORKFLOW.json separately
    (the matching combo workflow was not provided with the installer).
  - >-
    GGUF quant tier: the upstream installer prompts Q4_K_S (under 12GB,
    recommended) / Q5_K_S (12-24GB) / Q8_0 (24GB+). The manifest ships Q4_K_S
    active for both GGUF UNets; uncomment the matching tier to switch.
  - >-
    Combo pipeline: Qwen-Image generates the base image, then WAN 2.2 T2V A14B
    low-noise refines it. Both stages and their LoRAs/VAEs are installed here.
  - >-
    Use a Nightly ComfyUI build so the 8-step Qwen-Image Lightning LoRA and the
    GGUF UNet loader are supported.
post_install:
  - >-
    Restart ComfyUI (or use ComfyUI-Manager then Install missing custom nodes
    for any node Python deps).
  - >-
    If you hit "Torch not compiled with CUDA enabled", reinstall the CUDA torch
    build into the embedded/venv python, e.g.
    python -m pip install --force-reinstall torch torchvision torchaudio
    --index-url https://download.pytorch.org/whl/cu121
  - >-
    Load QWEN_COMBO_ULTRA_WORKFLOW.json (not bundled with this pack).