
MOSS-Image ships as two checkpoints, Base and SFT. What SFT changes, which one to pick for making images or training LoRAs, and how to run each in ComfyUI or Python.


Run MOSS-Image in ComfyUI: the four official templates, which model files to download and where they go, and the steps, CFG and SLG settings that matter.


Write MOSS-Image layout prompts by hand: the summary and items fields, yxyx bounding boxes in centered pixel coordinates, a converter, and examples to copy.


Write MOSS-Image prompts that render clean Chinese and English text: describe the finished image, quote the exact words and control line breaks. With examples to try.
