Back to search
VoxKite
MCP serverGitHub
Local speech recognition and synthesis for macOS: voice-engine library and voice-runtime MCP server
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/diana-random1st--voxkite)HTML
<a href="/listings/diana-random1st--voxkite"><img src="/api/listings/diana-random1st--voxkite/badge.svg" alt="findsafeskills safety rating" /></a>Install
i
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 35 lines
- # VoxKite
- VoxKite provides local speech recognition and synthesis for macOS. Use the
- `voice-engine` Rust library in an application, or run `voice-runtime` to serve
- `voice_speak` and `voice_listen` via MCP on localhost.
- ## Embed the engine
- Pin the Git revision of `voice-engine` and use the `transcribe-cpp-sys` patch
- from the same VoxKite revision. The patch retains the short-form encoder
- optimization used by the Whisper engine.
- - For STT, pass the local Whisper Large v3 Turbo GGUF to
- `NativeSttEngine::with_model_path`. The engine accepts mono 16 kHz PCM.
- - For synthesis with a reference voice, call `QwenTtsEngine::load_with_source`
- with the model source, the reference audio and its transcript. Use
- `QwenModelSource::LocalDirectory` or `QwenModelSource::HubCacheOnly` for
- local model files. Supply audio that you have permission to use.
- ## Build and run
- On Apple Silicon, from this checkout:
- ```sh
- cargo build --locked -p voice-runtime --bin serve
- ```
- The headless server uses `127.0.0.1:4525` by default and keeps its model files
- under `~/Library/Application Support/DianaVoice`. For a development check,
- see the fixture setup in [scripts/mcp-curl-gate.sh](scripts/mcp-curl-gate.sh).
- The check requests transcription of a WAV file and plays a synthesis sample.
- Diana.app provides microphone capture, the avatar and the desktop controls
- when it embeds VoxKite. The models and any reference recording are supplied
- by the operator.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.