Fish Audio launches Live 3, a full-duplex voice model that knows when to shut up
HeyAmit_ · x · 2026-11-11
Fish Audio released Live 3, a full-duplex voice model that listens while talking, produces natural backchannel cues like "mm-hmm," and knows when to stay silent. A demo on X showed Fish building a working voice agent named Jane (campaign deadline assistant) within 45 seconds of a single tweet. The product targets enterprise voice scenarios (drive-thru, retail support, flight changes, clinics), with free custom agent builds for the first 1,000 use-case submissions.
Related event: Fish Audio Launches Full-Duplex Voice Model Live 3 and Agent Platform(9 posts)→
More from Multimodal
- Seedream 5.5 lands on Lovart with RGBA transparency and better in-scene text rendering — aziz4ai · 2026-11-11
- Claude turns 27 reference images and a life allegory into a game with 1,159 paintings — gorkem · 2026-11-11
- rea-video: open-source tool reverse-engineers any video shot by shot, 6.5k GitHub stars — songguoxiansen · 2026-11-11
- DEZO: an AI-generated teaser for a 'lost' 1980s cyberpunk action film — RealBrainRocket · 2026-11-11
- Seedream 5 nails darkfield microscopy: AI diatoms with glass shells and chloroplasts — CurieuxExplorer · 2026-11-11
- Early MiniMax H3 Image test: CAD-style landmark redesigns — CurieuxExplorer · 2026-11-11