Muse: Text-To-Image Generation via Masked Generative Transformers

[Mactronn]

GoogleAI announced "Muse", a text-to-image generation/editing model via Masked Generative Transformers:

- Achieves new SOTA
- Zero-shot, Mask-free editing
- Zero-shot Inpainting/Outpainting
- 900M params

Paper: https://arxiv.org/abs/2301.00704
Project Page: https://muse-model.github.io/

Seems interesting..

 

[Steptoe]

Looks like huge leaps in speed and performance, but as it's a Google project it's doubtful we'll get to use it.

 

[Mactronn]

Looks like huge leaps in speed and performance, but as it's a Google project it's doubtful we'll get to use it.

Yes, seems to be 2x-3x faster than SD.