A quality, zero-shot lipsync pipeline built with MuseTalk, LivePortrait, and CodeFormer. You can run the latest, production-ready version of SieveSync here both through the playground and the API.
This repository contains the code for the pipeline we deploy to Sieve. It's meant for the developer community to have a nice way to experiment with lipsyncing pipelines. The community may find our implementation of face alignment and the pipeline architecture useful.
If you'd like to run this without Sieve, you may choose to swap out the model dependencies found at the top of main.py with MuseTalk, LivePortrait, and CodeFormer implementations.
You should then see a URL you can use to access your deployed function via the playground or API. As you make changes to the code, you can redeploy with the same command and you'll have an easy-to-use playground and production-ready API to test your changes.