NetEase Youdao's R2T2 + T3PO form a fully open-source real-time interpretation stack
thetripathi58 · x · 2026-09-23
A breakdown of how NetEase Youdao's two open-source models compose into a real-time speech translation stack:
- Confucius4-R2T2: low-latency, high-accuracy true streaming ASR with configurable 80ms–2s decode chunks and append-only output (committed text never gets revised), suited for live captioning, downstream LLM agents, and simultaneous interpretation. 487 stars on GitHub, ships WebSocket server/client examples.
- Confucius4-T3PO: a 14B streaming translation model that decides when to READ/WRITE and translates incrementally.
R2T2 turns speech into stable incremental text → T3PO translates it incrementally — interesting for real-time interpretation, meetings, and voice agents. Both have open repos, HF demos, and community GGUF quantizations for local deployment via llama.cpp/Ollama — not locked behind a hosted API.
Related event: NetEase Youdao's Open-Source Speech Models Top Hugging Face Trending(5 posts)→
More from Multimodal
- Endless screensaver of self-assembling quantum cryostats, made with Gemini 4 Carbon — hapticdata · 2026-11-12
- ADM 3 Cloud flops in image evals: ranks #86 of 107 in text-to-image — ArtificialAnlys · 2026-11-12
- ADM 3 Cloud ranks #47 of 80 at photo enhancement and restoration — ArtificialAnlys · 2026-11-12
- Apple's ADM 3 Cloud tested: #42 at image editing, #85 at text-to-image — ArtificialAnlys · 2026-11-12
- Object-level edits are ADM 3 Cloud's strongest skill, ranking #27 of 80 — ArtificialAnlys · 2026-11-12
- Apple's ADM 3 Cloud image model ranks #42 in editing and #85 in text-to-image on Artificial Analysis arena — ArtificialAnlys · 2026-11-12