GitHub radar
Open AI model that draws pictures with clean, readable text inside
MOSS-Image-SFT is an open image model that keeps text sharp and readable instead of blurry letters. You get a tool for covers and diagrams that runs on your own machine, no subscription needed.
It's an open model for creating and editing images with precise text inside — letters and objects are set through separate markup instead of being guessed by the network.
When it helps
Useful for covers, memes, and diagrams with captions; skip it if you need speed — one image takes a while to generate.
Pros
- Text comes out cleaner and more even than the base model
- Free for commercial use (Apache 2.0 license)
- Runs offline, no paid API or subscription
- Ready-made ComfyUI plugin (no-code image tool) — no coding needed
Cons
- Generation is slow — 50 steps, no turbo mode
- Needs a GPU with at least 8GB of video memory
How to set it up — step by step
- 1Open your AI agent (Claude Code, Codex) inside your project folder.
- 2Send the agent the text from the block below to install the model via ComfyUI or diffusers (library for running models).
- 3Wait for the agent to download the model (about 3GB compressed, up to 10GB in full quality).
- 4Check the result — a generated image with readable text, like a captioned cover.
- 5If the text looks off, ask the agent to describe it more precisely using JSON layout (structured instructions for what goes where).
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the model OpenMOSS-Team/MOSS-Image-SFT from Hugging Face for me: https://huggingface.co/OpenMOSS-Team/MOSS-Image-SFT — through ComfyUI or diffusers, whichever is easier on my hardware. Generate an image with readable text inside, for example a cover with a caption, and show me the result.
A regular laptop with a GPU of 8 GB or more - the compressed version of the model takes about 3 GB, the full-quality one about 10 GB.
Open on Hugging Face▌ More finds