comfyui-mcp/packs/ernie-combo/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: ernie-combo
display_name: ERNIE × Z-Image-Turbo Combo
family: ernie
kind: txt2img
description: >-
  Combo pipeline chaining ERNIE-Image Ultra with Z-Image-Turbo — one model's
  output feeds/refines the other ("ZIT → ERNIE", "ERNIE → ZIT", two-times
  combos). ERNIE brings precise multilingual TEXT RENDERING; Z-Image-Turbo adds
  fast, sharp detail. Sliced + un-bypassed from the ERNIE monolith's combo
  groups into a standalone graph. Heavier than the single-model packs — loads
  both models, so expect higher VRAM and longer runs.
vram: "12GB+"
workflow: workflow.json
skill: ernie-image
launch_args: []
external_models:
  - hirohiko-araki-style-ERNIE_000001250.safetensors
  - ernie-anime-v1.safetensors
sources:
  ernie: https://huggingface.co/Comfy-Org/ERNIE-Image
  z_image: https://huggingface.co/Comfy-Org/z_image_turbo
  docs: https://docs.comfy.org/tutorials/image/ernie-image/ernie-image
  mirror: https://huggingface.co/Aitrepreneur/FLX
notes:
  - Run the generated installer from your ComfyUI root (folder with custom_nodes/
    and models/).
  - Downloads BOTH model families (ERNIE + Z-Image-Turbo, ~32GB total). Use the
    ernie-txt2img pack instead if you only want plain ERNIE.
  - Sliced + un-bypassed from the monolith's COMBO groups; the opt-in prompt
    enhancer LLM is left off by default.
  - The z-image VAE is saved as z-image-ae.safetensors (its weights differ from
    Flux/ERNIE's ae.safetensors despite the same size — avoids a filename clash).
post_install:
  - Restart ComfyUI (or ComfyUI-Manager -> Install missing custom nodes for any
    node Python deps).
  - Load workflow.json; set the LoadImage node if the combo path you use takes a
    source image.