RynthV

A singing voice synthesis engine built with Rust and egui.

简体中文 · 日本語 · 한국어 · Русский

This fork adds macOS support. See macOS below and MACOS-SUPPORT.md for what changed and why. Upstream: cubeww/rynthv.

Building and running

Requires Rust 1.99.0, pinned by rust-toolchain.toml.

Windows

Needs MSVC, the Windows SDK and a GPU with DX12 support.

cargo run --release

Linux

Supports X11 and Wayland with Vulkan or OpenGL. Build example for Ubuntu 24.04:

sudo apt install build-essential pkg-config libasound2-dev libwayland-dev libxkbcommon-dev \
  fontconfig fonts-noto-cjk libvulkan1 libegl1 libgl1 xdg-desktop-portal xdg-desktop-portal-gtk
cargo build --release --locked --bins
./target/release/rynthv

Requires a working graphics driver, an ALSA audio device and a desktop XDG portal backend (xdg-desktop-portal-kde on KDE).

macOS

Needs macOS 11 or newer and a GPU with Metal support. On Apple Silicon the build is native arm64.

cargo build --release --bins
./target/release/rynthv

No extra system packages are required: Metal is the rendering backend, CoreAudio provides audio output, and the UI font is loaded from the system (PingFang SC, falling back to Hiragino Sans GB and then STHeiti).

Because the project pins Rust 1.99.0 in rust-toolchain.toml, rustup installs that toolchain automatically on first build.

Voice libraries and configuration

On first launch, select the voice library directory and the pronunciation dictionary directory in Settings. The program never probes resource paths automatically and ships without voice libraries, dictionaries or user model contexts.

Voice type	Windows	Linux	macOS
First-generation voices	Supported	Supported	Supported
Second-generation voices	Model context required	Model context required	Model context required

Configuration directories:

Platform	Configuration	Model context
Windows	%LOCALAPPDATA%/RynthV	%APPDATA%/RynthV/context.json
Linux	$XDG_CONFIG_HOME/RynthV (or $HOME/.config/RynthV)	context.json in the configuration directory
macOS	~/Library/Application Support/RynthV	context.json in the configuration directory

The command-line synthesis and diagnostic tools are documented under cargo run --release --locked --bin rynthv-synth -- --help.

Importing a model context on macOS

Second-generation voices need a model context. Three of the four sources work on macOS:

Source	Windows	Linux	macOS
Context file	Yes	Yes	Yes
Manual entry	Yes	Yes	Yes
Local session file	Yes	No	No
From the original editor (online)	Yes	No	No

The last two read the original editor's session file or process memory, which only exists on Windows x64. On macOS the UI disables those two sources and points at the context-file and manual-entry paths instead.

Platform support at a glance

Feature	Windows	Linux	macOS
Editor UI	DX12	Vulkan / OpenGL	Metal
Audio output	WASAPI	ALSA	CoreAudio
UI font	Microsoft YaHei	fontconfig (zh-cn)	PingFang SC / Hiragino Sans GB / STHeiti
Primary shortcut modifier	Ctrl	Ctrl	⌘ Command
Native CPU	x86-64	x86-64 / arm64	arm64 (Apple Silicon)

License

MIT. This fork keeps the original copyright notice; see LICENSE.