MOSS-Image: 5B Model for Text-Rich, Layout-Controlled Images
Published 31 October 2026
MOSS-Image is a 5B-parameter base DiT model for high-quality text rendering, structured layout control, long prompts, and anime or photorealistic image generation. The repository includes inference tools, ComfyUI support, layout-to-image workflows, and guidance for fine-tuning.
Read the original on github.com →