Welcome aMUSEd: Efficient Text-to-Image Generation
New model
Original post · huggingface.co ↗Summary by forck
Hugging Face introduces aMUSEd, an efficient non-diffusion text-to-image model based on Masked Image Modeling, as an open reproduction of Google's MUSE.
It is released as a research preview with a permissive license, integrated into the diffusers library, and features ~800M parameters with zero-shot inpainting capability.
