Float In ComfyUI With Wan VACE - Audio-driven Talking Portrait
Added 2025-05-23 13:53:34 +0000 UTC
Video Tutorial : https://youtu.be/YTZ5J3KcC60
For Patreon Supporters: https://www.patreon.com/posts/129689876
In this video, we explore Float, a cutting-edge AI talking portrait tool that generates lifelike lip-syncing avatars from just an image and audio input. Learn how to install and run Float in ComfyUI, optimize settings for facial expressions and emotions, and integrate it with WAN 2.1 Vace for advanced outpainting to create dynamic, full-body talking videos. Whether you're a content creator, AI enthusiast, or developer, this tutorial covers everything from basic setup to advanced workflows, including VRAM optimization and multi-GPU support for longer video generation.
Who is this for?
Content creators looking for realistic AI avatars
AI developers exploring lip-syncing frameworks
Tech enthusiasts experimenting with ComfyUI workflows
Video editors wanting to enhance talking-head productions
Why it matters:
Float offers smoother facial animations compared to older tools like Sonic and Latent Sync, while integration with WAN 2.1 Vace enables full-scene outpainting for professional-quality videos. This workflow is VRAM-friendly, making it accessible for mid-range hardware.
Resources:
Generative Motion Latent Flow Matching for Audio-driven Talking Portrait
https://deepbrainai-research.github.io/float/
https://github.com/deepbrainai-research/float
yuvraj108c/ ComfyUI-FLOAT
https://github.com/yuvraj108c/ComfyUI-FLOAT
Attached the example workflow for Float , you can practice and try it with other AI video like the example show in the video tutorial:
Comments
how I can fill with body like in YouTube clip
Pronprom Pakpaim
2025-07-09 15:48:43 +0000 UTC
