No Distillation Needed: Single-Pass Real-Time Talking Heads via Acausal Noise Shaping
Audio-driven facial animation underpins real-time avatars, telepresence, and embodied virtual agents.
ProofPaper ↗
Key points
- And it must run online: each frame emitted from audio observed up to the current time, at interactive rates.
- Audio-conditioned facial motion occupies a comparatively low-dimensional manifold, a regime where a single-pass GAN suffices.
- We show that a causal, time-invariant generator driven by i.i.d. noise cannot suppress its output spectrum over a band without collapsing its per-step innovation.
- FaceGAN emits expression and head pose in a single forward pass per frame and matches or outperforms state-of-art approaches in generation quality.
Sources (1)
- [1]No Distillation Needed: Single-Pass Real-Time Talking Heads via Acausal Noise ShapingarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 01:30 AM
Audio-driven facial animation underpins real-time avatars, telepresence, and embodied virtual agents.
And it must run online: each frame emitted from audio observed up to the current time, at interactive rates.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 8, 2026Normalizing Trajectory Models
- Oct 8, 2026LEGO: A Lifting-Free Approach for Exocentric-to-Egocentric Video Generation
- Oct 8, 2026One Block, Multiple Depths: Recurrent Vision Transformers with Depth-Programmed Experts
- Oct 7, 2026Enabling Preference-driven Unlearning in Few-step Distilled Text-to-Image Diffusion Models
- Oct 7, 2026Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning
- Oct 7, 2026MIMESIS: Learning User Simulators as Training Environments for Interactive Agents