https://huggingface.co/SeFi-Image/SeFi-Image-5B-RL

SeFi-Image is intended for research and creative text-to-image generation, including prompt following, bilingual text rendering, style exploration, and model development. The Base checkpoints are suitable starting points for fine-tuning and analysis. Turbo checkpoints are intended for fast generation. The RL checkpoint is intended for stronger alignment-oriented generation.


FamilyModelCheckpointStepsGuidancetime (285H)test v1test v2test v3
BaseSeFi-Image-1B-BaseSeFi-Image/SeFi-Image-1B-Base504.02m 18.09s 


BaseSeFi-Image-2B-BaseSeFi-Image/SeFi-Image-2B-Base504.03m 50.27s


BaseSeFi-Image-5B-BaseSeFi-Image/SeFi-Image-5B-Base504.08m 5.20s


RLSeFi-Image-5B-RLSeFi-Image/SeFi-Image-5B-RL504.08m 2.03s 
right arrowright arrow

TurboSeFi-Image-1B-turboSeFi-Image/SeFi-Image-1B-turbo41.0
right arrowright arrowright arrow

TurboSeFi-Image-2B-turboSeFi-Image/SeFi-Image-2B-turbo41.0



TurboSeFi-Image-5B-turboSeFi-Image/SeFi-Image-5B-turbo41.0