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),
39 lines (38 loc) • 1.82 kB
YAML
name: ideogram
display_name: Ideogram 4 (Ideogram Ultra)
family: ideogram
kind: txt2img
description: >-
Ideogram 4 (Ideogram Ultra) — an open-weight text-to-image model with
best-in-class text rendering (logos, posters, typography). Supports AREA
PROMPTING and structured JSON prompts via the Ideogram4 Prompt Builder KJ
node (ComfyUI-KJNodes), with a Gemma + Qwen3-VL text-encoder stack and the
FLUX.2 VAE. Great for posters, flyers, brand kits, and text/logo-heavy art.
vram: "~16GB"
workflow: workflow.json
skill: ideogram-ultra
launch_args: []
sources:
patreon: https://www.patreon.com/Aitrepreneur
model: https://huggingface.co/Comfy-Org/Ideogram-4
mirror: https://huggingface.co/Aitrepreneur/FLX
notes:
- Run the generated installer from your ComfyUI root (the folder containing
custom_nodes/ and models/).
- Area prompting / structured JSON prompts are driven by the
Ideogram4PromptBuilderKJ node from ComfyUI-KJNodes; rgthree-comfy provides
the Power Lora Loader used in the workflow.
- IMPORTANT — the V2 workflow loads TWO UNet models and applies the LoRA to
BOTH the normal (ideogram4_fp8_scaled) AND the conditional/unconditional
(ideogram4_unconditional_fp8_scaled) UNets. Keep both downloads and the LoRA
wired to both UNETLoaders or guidance will break.
- The fp8_scaled weights are the official Comfy-Org Ideogram-4 release; the
text encoders and VAE are mirrored on Aitrepreneur/FLX.
post_install:
- Restart ComfyUI (or use ComfyUI-Manager → Install missing custom nodes for
any node Python deps).
- Install the JSON prompt template pack — copy the 25 templates from this
pack's templates/ folder into
ComfyUI/user/default/kjnodes/ideogram4/templates so the Ideogram4 Prompt
Builder can load them.
- Load workflow.json.