๐ค MOSS-Image is now open source on
@huggingface .
Built on a 5B Diffusion Transformer (DiT), MOSS-Image brings Chinese & English text rendering and controllable layouts to image generation โ from posters and infographics to creative visuals.
@Open_MOSS
What it offers:
๐ Long prompts & text rendering โ up to 4K-token prompts, including multi-line text
๐ Experimental JSON layout control โ specify text, objects, and their positions using bounding boxes
๐ผ๏ธ Flexible generation โ multiple aspect ratios, resolutions, and customizable inference settings ๐ ๏ธ Developer workflows โ Python (Diffusers), Gradio, ComfyUI, and a Prompt Enhancer Skill
๐ฆ Base & SFT checkpoints are both available, with open model weights and Apache-2.0 licensed inference code.
๐ Huge thanks to
@sgl_project @vllm_project @ModelScope2022 for the Day-0 support!