HF speech-to-speech merges NVIDIA Parakeet Unified STT backend
andimarafioti · x · 2026-09-21
The Hugging Face speech-to-speech project (13.3k stars) merged a PR adding NVIDIA Parakeet Unified as an optional STT backend, addressing issue #235.
Details:
- Enabled via --stt parakeet-unified, default model nvidia/parakeet-unified-en-0.6b
- Shared NemoASRSTTHandler added, with optional nemo extra
- Each VAD utterance is transcribed with NeMo ASRModel.transcribe; no cache-aware chunked decoding added
- parakeet-tdt remains the default; missing extra fails at setup and points to speech-to-speech[nemo]
- mypy and ruff checks all pass
More from coding & agent
- AI Comes for the If Statement: Tomasz Tunguz on AI replacing hardcoded rule logic — hardimanjames · 2026-09-22
- MrKan turns AI agents into kanban teammates for managing multiple projects with guardrails — agf2007 · 2026-09-22
- Thorsten Ball: Code Review Will Die, and Unit Tests Might Follow — rseroter · 2026-09-22
- Gemini 6 Pro Suddenly Lost Ability to Audit Vercel-Deployed Projects — aklibf · 2026-09-22
- Tinfield 1 open-weight model claims to beat Claude Opus 4.8 on Terminal-Bench — victormustar · 2026-09-22
- Qwen Code Desktop v0.24.3 ships bwrap sandbox, DingTalk output and token budgets — github-actions[bot] · 2026-09-22