QuixiVoice: a Rust voice stack for any agent, from VAD to TTS in one pipeline
QuixiAI · x · 2026-11-16
Developer QuixiAI is advancing QuixiVoice, a voice stack for any agent combining a fast voice server and an MCP server, written in Rust with the burn framework.
The pipeline chains together LAION's 6k-diverse-reference-voices dataset, Silero VAD v5 for voice activity detection, NVIDIA Nemotron 3.5 ASR Streaming 0.6B for streaming speech recognition, Google Gemma 4 E2B IT (QAT) as the LLM, and Ming 3.0 TTS 0.5B for synthesis.
It currently runs on Metal kernels, with ports to CUDA, CDNA and XPU planned for broader hardware coverage.
More from coding & agent
- Spend tokens on verification, not generation: poteto's talk on shipping quality, not slop — gnukeith · 2026-11-16
- Alibaba open-sources open-code-review, a hybrid AI code reviewer with 52k GitHub stars — abhishek__AI · 2026-11-16
- Gemini agent's Auto permissions mode asks per-file approval instead of per-directory — BenBlaiszik · 2026-11-16
- Musk shows Grok Bot auto-editing 23-hour Starship livestream into highlight reel — elonmusk · 2026-11-16
- Open-source CAD SolidCraft 0.42 ships KiCad import and MCP support for AI agents — philfung · 2026-11-16
- Token-efficient main-session + Thread agent architecture: weak router, strong workers, smart caching — dotey · 2026-11-16