YoVoice
leemysw/yovoiceReviewed by AI
Checked by RepoPicks Radar on
What is YoVoice?
yovoice is a desktop studio for voice and music that works on your own machine. It reads your text aloud in a built-in voice, copies a voice from a short recording, dubs a story line by line with a character per line, writes songs from a style and your lyrics, and arranges scores you can export as MIDI or separate tracks. There are installers for Windows, macOS on Apple Silicon and Linux; you choose and download the models inside the app. An optional panel chats with DeepSeek, OpenAI, Anthropic or a local Ollama to write lyrics and start the work for you.
YoVoice is free and open source (Apache-2.0) and can replace ElevenLabs, Suno, and Udio.
How easy is it to install YoVoice?
B · installer plus one simple step · Windows
- Windows B
- macOS B
- Linux B
Windows is the easiest route and sets the grade: run the setup .exe (a plain 78 MB one, or a 610 MB / 785 MB build that already carries the NVIDIA CUDA runtime), then do the one extra step the Quick Start asks for, "Download a model in Settings" — that is a B. macOS is the same with the .dmg, but only macOS 14 or newer on Apple Silicon. Linux has a .deb that also pulls in WebKitGTK and runs on the processor out of the box, with the note that "The Linux app doesn't update itself yet".
YoVoice on GitHub
- Stars
- 2,158on Nov 10, 2026
- Downloads
- 30,1862,600 this week
- Growth
- +96stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Nothing extra
- To build from source
- Go
- Disk space
- 785 MB–2.3 GB · from the installer
- Accounts or keys
- DeepSeek, Qwen, Kimi, OpenAI, Anthropic or a local Ollama, only for the optional assistant panel
- Cost
- Free · The program is free under Apache-2.0 and the models run on your machine, so there is no per-character or per-minute bill. The optional assistant panel uses a key you bring yourself, and some model weights carry their own limits (see the signals).
- License
- Apache-2.0 (Permissive)
- Can replace
- ElevenLabs, Suno, Udio, Murf
- Last change on GitHub
- Nov 10, 2026
YoVoice risk signals (4)
- Attentioncopies a real persons voice
The README offers "zero-shot voice cloning" and the feature "clone a reference voice, design a voice from a description, adjust emotion and delivery". A short recording of someone is enough to make that voice say anything you type, and using a person's voice without their permission is restricted in several countries.
AI review
- Attentionmodels are non commercial only
The Supported Models section says "OmniVoice, YuE2 and Breeze TTS 2 weights are restricted to non-commercial use; Stable Audio 3 is free below USD 1M annual revenue under the Stability AI Community License", and the licence note adds that "IndexTTS models are covered separately by their model license". The Apache-2.0 badge covers the program, not what every model lets you sell.
AI review
- Infomodels are downloaded after install
"All models are downloaded inside the App or with the CLI", on top of installers that already reach 785 MB, and a Vulkan GPU runtime that is fetched in Settings on Linux. The space the app ends up taking depends on how many voice and music models you add.
AI review
- Infocan serve other machines over the network
"Run yovoice serve to provide an authenticated HTTP API and MCP inference service for other machines." The README says that service asks for authentication; it is off until you start it.
AI review
Evidence is shown in its original language.
What Radar could not check about YoVoice
- Whether generation is fast enough on a machine with no NVIDIA card, since Linux runs on the processor out of the box
- How much disk the app really takes once several voice and music models are downloaded
- Whether the voices and songs come out like the demo page shows
- Whether the Vulkan runtime offered in Settings works with your graphics card on Linux
- What the optional assistant panel sends to the provider you choose