OpenMOSS-Team/MOSS-Image-SFT — new model trending #30 on Hugging Face
OpenMOSS released MOSS-Image-SFT, a fine-tuned text-to-image model supporting prompts, tags, and JSON layouts.
OpenMOSS published MOSS-Image-SFT, a supervised fine-tuned text-to-image model derived from MOSS-Image-Base and trending on Hugging Face. It keeps the same architecture and inference pipeline and supports natural-language prompts, tag sequences, and JSON layouts for text placement and object arrangement. Recommended sampling uses a flow-matching Euler sampler, 50 steps, and CFG of 3, with a bundled FLUX.2 small VAE decoder. Weights, a technical report, a demo Space, and a ComfyUI plugin are available.
46