# yovoice

> yovoice is an open-source, local-first audio and music creation studio for macOS, Windows, and Linux. It creates expressive speech, cloned voices, full vocal songs, instrumentals, sound effects, and multi-track scores on the user's own computer.

Current release: v0.5.0.

## Create

- **Story dubbing and speech**: Import text, EPUB, or subtitle files, split them into lines, assign characters, generate line by line, and edit the result on a timeline with automatic audio checks and target loudness export.
- **Voice and expression**: Clone a reference voice, match a reference performance, design a voice with text, blend preset voices, or adjust emotion and delivery. Available controls vary by model.
- **Music and sound effects**: Use ACE-Step 1.5 Turbo / XL Turbo for songs or instrumentals, YuE2 3B for vocal songs and score-based instrumentals, or Stable Audio 3 for instrumental music and short sound effects (1–30 s). Supports rework, inpainting, and word-by-word lyric alignment.
- **Score**: Write parts by bar and beat, import MIDI, or let an Agent compose. Render with MuseScore General, GeneralUser GS, or SGM-V2.01; export MIDI and stems. Score rendering needs no inference engine or GPU and is exactly repeatable.
- **Optional AI services**: DeepSeek, Qwen, Kimi, OpenAI, Anthropic, Ollama, and compatible APIs can write scores, song styles, lyrics, and titles. Keys stay on the computer; request content goes to the selected service. Other features work without an AI service.

## Run and integrate

- The desktop app supports macOS 14+ on Apple Silicon, Windows 10/11 x64, and Linux x64 (Debian/Ubuntu). It uses audio.cpp v0.9.3, Metal on macOS, CPU / NVIDIA CUDA / experimental Vulkan on Windows, and bundled CPU / Vulkan GPU on Linux. CUDA 12.4 and 13.3 installers are available.
- Model inference and score rendering run locally after the required models or sound fonts are downloaded. Projects, voices, models, and settings are stored in the local data directory.
- The standalone CLI and Agent Skill support scripted speech generation and score rendering. Run `yovoice serve` to provide an authenticated HTTP API and MCP inference service for other machines.
- The website's audio samples are pre-generated speech files. The browser does not run a model.

## Product resources

- [yovoice homepage](https://yovoice.leemysw.com/): Product overview, interactive speech samples, and download links.
- [Models and licenses](https://yovoice.leemysw.com/models.md): Model capabilities, parameters, file sizes, sound fonts, hardware guidance, and license notes.
- [Complete public guide](https://yovoice.leemysw.com/llms-full.txt): Expanded product facts and operating constraints.

## Developer resources

- [Agent Skill](https://github.com/leemysw/yovoice/tree/main/skills/yovoice): CLI and agent workflow instructions.
- [CLI guide](https://github.com/leemysw/yovoice/blob/main/docs/cli.md): Local speech, score, and MCP workflows.
- [API / MCP deployment](https://github.com/leemysw/yovoice/blob/main/docs/api.md): Authenticated local and remote services.
- [Source repository](https://github.com/leemysw/yovoice): Source code, releases, and issue tracker.
- [Downloads](https://github.com/leemysw/yovoice/releases): Desktop and CLI releases.