Ultra-high-definition image demoiréing (UHDM) – New model that can remove the moire pattern you get when taking a photo of a screen. Project page, Github, paper.
Compositional human-scene interaction synthesis with semantic control (COINS) – Given a 3D scene and a text instruction, COINS can generate realistic human interaction with that scene. Project page, Github.
Counterfactual Image Manipulation via CLIP (CF-CLIP) – A new method for accurately editing StyleGAN-generated images with CLIP. Github, Paper.
It's one year to the day since I got my first batch of good CLIP Guided Diffusion outputs! I wrote the public notebook, "CLIP Guided Diffusion HQ 256x256", the next day. The prompt was "Desert dynamism by Giacomo Balla".
If you have anything you’d like to be featured or want to get in touch, give me a shout on Twitter or via email. Please also consider supporting me on Patreon so that I can spend more time creating content like this.