GitHub radar
Draw pictures with readable text inside, way faster
Kromoss is a retrained version of an open image model that draws and edits pictures while keeping the text inside them readable instead of turning letters into mush, and with its speed boost it can do that in 8 steps instead of 50.
It's a fine-tuned version of the open MOSS-Image model: it draws pictures from a text prompt and keeps any writing on them crisp.
When it helps
Good for covers, diagrams, memes, and anything with text baked into the image; not yet a go-to tool for serious production image work.
Pros
- Text inside pictures stays readable instead of turning to mush
- With the speed-up LoRA it draws in 8 steps instead of 50
- Drop-in file swap if you already run MOSS-Image in ComfyUI
- On the authors' own test, the fast version beats the slow original
Cons
- The fast mode trades a bit of flexibility for speed
- Needs a GPU with at least 8GB, or it won't run
- Full version is about 10GB, the compressed one about 5GB
How to set it up — step by step
- 1Open your AI agent (Claude Code, Codex) in your ComfyUI project folder.
- 2Send the agent the text from the block below.
- 3If MOSS-Image is already set up, swap in Kromoss and turn on the 8-step speed-up.
- 4If ComfyUI isn't set up yet, ask the agent to install it from scratch from huggingface.co/lodestones/Kromoss.
- 5Check the result: generate a picture with text and confirm the letters are crisp.
- 6Make sure your laptop's GPU has at least 8GB, or the model won't run.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the Kromoss text-in-image model for me: https://huggingface.co/lodestones/Kromoss — if I already have MOSS-Image in ComfyUI, swap the file for this one and turn on the 8-step accelerator so images draw faster.
A regular laptop with a GPU of 8 GB or more - the compressed build of the model takes about 5 GB, the full one about 10 GB.
Open on Hugging Face▌ More finds