d.
Single-pass GAN matches diffusion quality for real-time talking heads at 43 ms latency
Acausal noise shaping overcomes a fundamental spectral limitation of causal generators, enabling single forward pass without iterative sampling.
Industry
Yu Han · Dejan Markovic · Alexander Richard · Wojciech Zielonka · Akshay Venkatesh · Cheng-hsin Wuu · +1 more
Meta Reality Labs
Research Digest··2 min read
Authors from Meta Reality Labs present FaceGAN, a single-pass GAN for audio-driven facial animation that runs at 43 ms latency and 12x real-time on one GPU.
Why this paper
From Meta Reality Labs
In one line
A single-pass GAN with acausal noise shaping matches diffusion quality for real-time audio-driven facial animation at 43 ms latency.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors (3 noted)
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§