PKU's Motion-Omni generates speech and full-body motion end-to-end, faster than realtime
aigclink · x · 2026-09-07
Peking University released Motion-Omni, an end-to-end conversational system that generates speech together with synchronized full-body motion — facial expressions, hands, upper and lower body.
Key facts:
- RTF 0.78, total response time 4.32s, faster than realtime
- Motion is natively emitted from the speech generator's internal hidden states, so audio and motion are inherently synchronized — not a cascade of TTS then audio-driven motion
- 5.4x faster than the LOM teacher-cascading approach on the same audio
- Still offline (waits for the full user utterance), not streaming
The work is a step toward embodied AI avatars, moving digital humans and virtual streamers from a "prerecorded" feel toward live interaction.
Related event: PKU Releases Motion-Omni for Real-Time Speech-Motion Generation(3 posts)→
More from Embodied
- WanGP v18.81 adds Tencent's Hy-Nomas world model, playable on 8GB VRAM — cocktailpeanut · 2026-10-27
- MS-Human-700: open-source human model with 700 muscle actuators lands in MuJoCo — yisongyue · 2026-10-27
- NVIDIA open-sources FLORA, a robotics data engine ~125x cheaper than GPT-6 Astra — nvidia · 2026-10-27
- Embodied AI startups need a different operational discipline, investor says — ___Mufasaa · 2026-10-27
- π0.5 study: raising demos from 10 to 438 lifts Place Ball in Bowl success from 7% to 83% — stingning · 2026-10-27
- GPT-6 Astra controls dexterous hands zero-shot at just 5.2% success in 500 trials — stingning · 2026-10-27