Real-Time-Latent-Consistency-Model

There was a webcam demo a little while back that was pretty cool. It's mostly that the right-hand side seems to only mildly follow the patterns drawn on the left hand side. It still seems sort of useful but once it's running GAN speeds (30-60fps) and adheres more strongly to the input is when people will find it both useful and a joy to work with. Based on my general experience with text-to-image stuff, I assume this lack of adherence isn't always the case? Maybe another demo could show what it's like under ideal conditions. edit: Webcam demo https://github.com/radames/Real-Time-Latent-Consistency-Mode... It does however exhibit similar issues but the realtime constraints of live video make it quit interesting.

yes, just publish it on the readme.me here https://github.com/radames/Real-Time-Latent-Consistency-Mode...

Hi, I've built the demo. Unfortunately, it's running on a single GPU and can only be used concurrently by a few users. For a better real-time experience, you would need a dedicated machine. You can get the source here and how to run: https://github.com/radames/Real-Time-Latent-Consistency-Mode... See video here: https://twitter.com/radamar/status/1718783886413709542 or on the Github readme. FYI, this is made possible due to a new technique: https://latent-consistency-models.github.io, fine-tuning an existing models. The author will soon publish the training script. We'll see all the cool image models running at this speed! I'm excited for this! I can see many interesting experiments and projects emerging from this.