NVIDIA open-sources PersonaPlex-7B, a full-duplex voice model that listens and speaks simultaneously
VeryWellVersed · x · 2026-10-11
NVIDIA has open-sourced PersonaPlex-7B, a full-duplex conversational voice model that can listen and speak at the same time. Using a dual-stream architecture, it processes input and responds in real time instead of waiting for the user to finish, removing a major friction point in voice AI. The model is fully open-source and free.
More from Multimodal
- Video Prompt Craft: Splitting 30 Seconds into Timed Beats with Camera Notes — lmoroney · 2026-10-12
- Hook viewers with Magnific One: familiar-but-striking starting images — techhalla · 2026-10-12
- Animate cheap in Seedance 2.5 draft mode, then flip to HD without losing detail — techhalla · 2026-10-12
- Magnific Desktop makes videos so real people ask "wait, is this AI?" — techhalla · 2026-10-12
- One Prompt, A Full Map-Explainer Video: How A Creator Automated a YouTube Channel — itsOmSarraf_ · 2026-10-12
- Educator with RTX 5090 Asks How to Start ComfyUI for Science Shorts — AngryManzUK · 2026-10-12