comfyui-mcp/packs/ideogram-img2img/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: ideogram-img2img
display_name: Ideogram 4 — Image to Image
family: ideogram
kind: img2img
description: >-
  Ideogram 4 (Ideogram Ultra) image-to-image, running LOCALLY from the Comfy-Org
  fp8 DiT weights (NOT the Ideogram cloud API). Encodes an input image
  (LoadImage -> VAEEncode) and re-renders it under denoise control, guided by a
  Qwen3-VL text encoder and a Gemma4 prompt-builder LLM (JSON Prompt Builder
  subgraph) with a dual conditional/unconditional UNet guider. Single-purpose
  img2img graph sliced from the ideogram monolith (one pipeline, no group
  toggles), with KJ resize/sharpen post-processing and an off-by-default Power
  Lora Loader.
vram: "<8GB"
workflow: workflow.json
skill: ideogram-img2img
launch_args: []
sources:
  model: https://huggingface.co/Comfy-Org/Ideogram-4
  mirror: https://huggingface.co/Aitrepreneur/FLX
notes:
  - Run the generated installer from your ComfyUI root (folder with custom_nodes/
    and models/).
  - Local weights only — this does NOT call the Ideogram cloud API; the fp8
    Ideogram-4 DiT is loaded via UNETLoader and sampled with SamplerCustomAdvanced.
  - Sliced + un-bypassed from the monolith's "IMAGE TO IMAGE" group.
  - Set the input image in the LoadImage node and tune the denoise (PrimitiveFloat)
    to control how far the result departs from the source.
post_install:
  - Restart ComfyUI (or ComfyUI-Manager -> Install missing custom nodes for any
    node Python deps).
  - Load workflow.json.