GitHub radar
Draw and edit images on your own graphics card, no server
PrunaAI repackaged the P-H3-Image image-generation model, built on MiniMax-H3, to run directly in ComfyUI (a free visual tool for running image generation on your own computer) — one 24 GB gaming graphics card instead of a multi-GPU server.
It's a model that draws images from a text prompt and edits photos, using up to 8 reference images at once.
When it helps
Useful if you want to generate and edit images on your own machine instead of paying for a subscription and are willing to judge quality yourself on your own prompts; skip it without a 24 GB graphics card.
Pros
- Smallest weight set is a 20 GB download that fits a 24 GB card
- Renders an 8-megapixel image in just 8 steps
- Edits photos using up to 8 reference images at once
- ComfyUI reads the files directly, no manual file conversion
Cons
- Output quality wasn't verified — test it on your own prompts
- MiniMax-H3 license restricts use in the EU, UK, Korea and US without a separate license
How to set it up — step by step
- 1Open your AI agent (Claude Code, Codex) inside your ComfyUI project folder.
- 2Send the agent the text from the block below.
- 3Give the agent GPU access and let it download the weights (up to 20 GB).
- 4Check the agent picked the weight set that fits your graphics card (needs 24 GB VRAM).
- 5Ask the agent to generate your first image from a short text prompt.
- 6Look at the result and judge quality yourself — the spec doesn't guarantee it.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the P-H3-Image package for ComfyUI from PrunaAI: https://huggingface.co/PrunaAI/P-H3-Image-Comfy — pick the smallest weight set for my GPU, install it and show me how to generate my first image.
Needs a GPU with 24 GB of VRAM — the smallest weight set peaks at 21 GB of VRAM.
Open on Hugging Face▌ More finds