Creator tool · MIT · development
Shadowfetch Voice Studio
A private, local-only voice cloning studio for Linux. Record or import a clean reference clip, review the waveform and transcript, save a reusable voice, then turn a script into speech on your own machine. The repository states there is no telemetry and no cloud inference; the only network use is downloading model weights you approve.
This is not a cloud voice API and not part of the Linux ISO.
The repository aims at NVIDIA GPU machines; it does not claim every Linux desktop can generate speech.
- License
- MIT
- Platforms
- Linux (NVIDIA GPU aimed)
- Status
- Latest GitHub release v0.2.0
- Source
- Open on GitHub
- Category
- Creator Tools
- Engines documented in the README: Qwen3-TTS 1.7B Base and optional Chatterbox-Turbo.
- Transcription uses faster-whisper, CPU by default.
- Offline mode is described as blocking every request after you choose it.
Linux package from the tree
The README's supported Linux install is a .deb produced by scripts/build.sh, with the Python engine runtime created under the data directory on first run.
See scripts/doctor.sh, scripts/bootstrap.sh, and scripts/install-linux.sh in the repository. AppImage output is documented as optional and best-effort.