Audio-Visual Driven Compression for Low-Bitrate Talking Head Videos
This work addresses bandwidth-constrained scenarios for talking head video applications, representing a strong specific gain in compression efficiency.
The paper tackles the problem of low-bitrate talking head video compression by proposing an audio-visual driven codec that integrates 3D motion features and audio signals, resulting in a 22% bitrate reduction compared to VVC and 8.5% over state-of-the-art learning-based codecs while improving lip-sync accuracy and visual fidelity.
Talking head video compression has advanced with neural rendering and keypoint-based methods, but challenges remain, especially at low bit rates, including handling large head movements, suboptimal lip synchronization, and distorted facial reconstructions. To address these problems, we propose a novel audio-visual driven video codec that integrates compact 3D motion features and audio signals. This approach robustly models significant head rotations and aligns lip movements with speech, improving both compression efficiency and reconstruction quality. Experiments on the CelebV-HQ dataset show that our method reduces bitrate by 22% compared to VVC and by 8.5% over state-of-the-art learning-based codec. Furthermore, it provides superior lip-sync accuracy and visual fidelity at comparable bitrates, highlighting its effectiveness in bandwidth-constrained scenarios.