Shizuku-Duplex: Japanese full-duplex voice model built for AI characters debuts with demo
Scobleizer · x · 2026-11-08
Shizuku AI Labs has introduced Shizuku-Duplex, which it claims is the world's first full-duplex audio synthesis model built specifically for AI characters, with a Japanese-language conversation demo now public.
- Key capability: full-duplex interaction — the model can listen and speak simultaneously rather than in strict turn-taking, aiming at more natural human-AI dialogue
- Positioning: designed for AI character/virtual persona use cases rather than general-purpose assistants
- The updated Shizuku character demo showcases noticeably improved Japanese conversational flow
More from Multimodal
- Dev shows Astra (high) emulating PS1-style King's Field-like dungeon graphics — Dimillian · 2026-11-08
- Prompt share: hyperrealistic 3D celebrity caricatures with exaggerated heads and lifelike skin — azed_ai · 2026-11-08
- 2D-to-3D AI conversion turns stylized cartoon characters into photoreal renders — aziz4ai · 2026-11-08
- Scale invariance may fix why autoregressive image generation lags diffusion models — burkov · 2026-11-08
- Combining Pika LoRA and NVIDIA Open 2K for Free 2K Video Generation on Colab — lufeixr · 2026-11-08
- Free LTX-2.5 ComfyUI Workflow Pack: Text-to-Video, Image-to-Video, and Reference Sheets — Beneficial-Coast5460 · 2026-11-08