Sonix is a DAW for Linux. It records from the microphone, arranges on the timeline, mixes with real meters and exports what you hear. The audio engine runs on cpal and ALSA, projects are saved to disk, and nothing leaves your machine unless you connect an AI service yourself.
The ready-made binary needs neither Rust nor a build step. If you want the newest code, build from source; it takes a few minutes.
x86-64, glibc 2.35 or newer. Unpack and run. The archive has the binary, the manual, the changelog, the licences and a sha256 checksum.
curl -LO https://github.com/alexwest1981/sonix/releases/download/v0.9.0/sonix-0.9.0-x86_64-unknown-linux-gnu.tar.gz
sha256sum -c sonix-0.9.0-x86_64-unknown-linux-gnu.tar.gz.sha256
tar xzf sonix-0.9.0-x86_64-unknown-linux-gnu.tar.gz
./sonixAsks for a language, checks dependencies, builds with Cargo and adds a launcher to the application menu.
bash <(curl -fsSL https://raw.githubusercontent.com/alexwest1981/sonix/master/install.sh)If you already have a clone, bash install.sh in that folder is enough. After changing the code: bash install.sh --refresh.
The same binary, packaged in the omapak catalogue. The app's configuration and library live in the sandbox under ~/.var/app/.
unzip is only needed by the Suno import.The release is from 11 September 2026. Since then master has gained microtonal tunings, a modulation matrix, a groove pool and four stretch modes, among other things. They are in the source, not in the binary. To try them, run the installer alongside it.
Every line here connects to the audio engine or the project state. The details come from the app's own status list.
Drag a clip's edges to extend, loop or trim it. The slice tool cuts, automation draws curves, and markers split the song into sections.
A 16-step sequencer with its own channels, and a piano roll with note lengths, velocity, scales and recording from a MIDI keyboard.
Faders, peak meters, solo, mute, pan and width per track, plus buses, sends and a master chain.
Several takes per track, comping and a harmonizer. Pitch editing runs offline on the recorded file, so what you hear is what gets saved.
Split a mix into vocals, drums, bass and other with DSP in the standard build. The neural path is opt-in and downloads its model first.
Rust and egui, MIT or Apache-2.0. No account, no subscription, no telemetry. Seven interface languages, switched live, all complete.
See the code ↗The images come from the app itself. All 38 are in the screenshots folder.
It is an early release, and you may as well know the limits before you download it.
--features plugin-host. LV2 exists only in the catalogue, and plugin windows need a real X server.