VoxShift
github.com
- Free plan
- Windows
- macOS
- Linux
What VoxShift is
VoxShift is a free, MIT-licensed desktop app that changes microphone audio in real time with Retrieval-based Voice Conversion. It runs on Windows, macOS, and Linux for voice chats, games, and video calls without requiring code. The app lists six voices, a community marketplace, and controls for pitch, formant, index ratio, input and output gain, monitoring, and noise suppression. Users can train a voice from WAV, MP3, FLAC, or OGG files by selecting training epochs and a pitch-extraction method. Audio processing stays on the user’s computer. The project says it has no telemetry, analytics, crash reporting, or user accounts; network calls are limited to marketplace metadata and optional GitHub updates. The README lists under 200 ms latency on a modern GPU and requires PyTorch with CUDA, ROCm, or Metal support for real-time inference. Virtual audio routing is described for Discord, OBS, and Zoom. The repository has no available releases.
Who it suits
VoxShift suits people who want local AI voice conversion for chats, games, or calls without writing code. It requires compatible GPU acceleration for real-time inference.
Good to know, and what to watch
- Free and open source
- Local processing with no telemetry
- Advanced voice controls
- Voice training from common audio files
- Windows, macOS, and Linux support
- Real-time use requires supported GPU acceleration
- Repository has no available releases
- Free trial is not offered
Plans
Every plan and priceFree
Free plan
The verdict
VoxShift offers private, local voice conversion with training, marketplace models, and virtual-audio routing. Users should check GPU support and the lack of published releases before setup.
What the maker publishes
- Audio controls
- Listed controls include pitch shift, formant shift, index ratio, input and output gain, monitor mode, and noise gate or suppression. source
- Download availability
- The repository's Releases page states that there are no releases available. source
- Hardware requirement
- The README says real-time inference requires PyTorch with CUDA, ROCm, or Metal support for GPU acceleration. source
- Integrations
- The README describes routing processed audio through a virtual cable for use as a microphone in Discord, OBS, and Zoom. source
- Intended audience
- The README presents VoxShift for people who want to use AI voice conversion in voice chat, games, or video calls without writing code. source
- Latency
- The README lists under 200 ms end-to-end latency on a modern GPU. source
- License
- The repository identifies the project as MIT licensed. source
- Local processing
- The project says audio processing happens on the user's machine and audio never leaves the computer. source
- Network calls
- The README says network calls are limited to the marketplace API for metadata and opt-in GitHub auto-updates. source
- Privacy
- The README says VoxShift has zero telemetry, no analytics or crash reporting, and no user accounts. source
- Purpose
- VoxShift is a desktop app that transforms microphone audio in real time using Retrieval-based Voice Conversion (RVC). source
- Virtual audio drivers
- The README lists VB-Audio VB-Cable for Windows, BlackHole 2ch for macOS, and a PipeWire virtual sink for Linux. source
- Voice marketplace
- The README describes browsing and one-click downloading community voice models from voice-models.com, with categories including anime, masculine, feminine, character, and robotic. source
- Voice training
- The README says users can train a voice by dropping WAV, MP3, FLAC, or OGG files and choosing training epochs and a pitch extraction algorithm. source
Checked against the maker's pages on 7 Oct 2026.