NetEase Youdao open-sources Confucius4-R2T2 and T3PO, a full real-time voice AI stack
thetripathi58 · x · 2026-09-23
NetEase Youdao open-sourced two models that hit #1 on Hugging Face Trending in their categories: ASR model Confucius4-R2T2 and 14B streaming translation model Confucius4-T3PO.
- R2T2 is built for voice agents: committed transcript chunks become permanent, new text is only appended, never rewritten — so "book" flipping to "look" can't derail a search or booking. Stable input equals reliable voice AI
- T3PO decides when to READ/WRITE over R2T2's stable incremental text and translates incrementally; audio is processed in slices as small as 160ms (streaming step, not end-to-end latency)
- Together they form a fully open-source real-time voice/interpretation stack for live interpretation, meetings, and voice agents: your agent can start understanding while the user is still speaking
Related event: NetEase Youdao's Open-Source Speech Models Top Hugging Face Trending(5 posts)→
More from coding & agent
- User automates entire video clipping workflow with Nous Research's Hermes agent — Teknium · 2026-11-12
- As models improve, writing good evals gets harder and more important — amankhan · 2026-11-12
- Prime agent ships with day-one multi-agent communication, runs in Minecraft unmodified — sarahcat21 · 2026-11-12
- Autonomous agent edits just 1,024 weights to lift Qwen3.5 0.8B from 7.4% to 37.4% on MATH-500 — VictorSanh · 2026-11-12
- Omarchy 4.6 lets you share local AI models across your tailnet, says DHH — alexcovo_eth · 2026-11-12
- Coding harnesses hand shell access to inference providers — 4-step gateway mitigation — shensi · 2026-11-12