Pinned
Reve 2.0 is here! Ever since we moved away from the latent autoencoder, we are making yet another bet to move away from the pure text-input for image generation!
Songwei Ge
@Songwei_Ge
Joined September 2020
- Making image generation work with non-textual input allows more precise rendering of our imagination! I have been fascinated by this idea since my PhD. It's really nice to see it comes together so well with Reve through layout!
- We are shipping Reve 2.1 today! Our model got better again in only one month after Reve 2.0 , ranked at #2 in @arena and @DesignArena! Better, more aesthetic and intelligent models are always coming soon from @reve.
- I'm beyond excited about this native 4k image generation model that we've worked on for months! It feels great that the image can be directly used as a wallpaper.Reve's new text-to-image model is here. Really proud of the team to rank at #3 with the big labs. To my knowledge, we are the first lab to use native pixel space diffusion without latent autoencoder at 4k (16MP) resolution for production level image generation.
