Skip to content

Usage

Starting the application

Installed with the Windows installer

Search for "Yaas" in your installed applications and start it.

Installed with the macOS installer or the Linux AppImage

Launch it like any other application (from /Applications on macOS, or by running the .AppImage file on Linux).

Installed with uv / pip

Activate your virtual environment and run:

yaas

Extracting stems

  1. Use the embedded browser to search for or navigate to the video you want. A playlist page (youtube.com/playlist?list=...) works too: all its videos are split, one after the other.
  2. Click Start.
  3. Wait — separation can take a while, especially on CPU. The progress bar follows the download, the model download on first use, and the separation itself.
  4. When the job is done, the status log lists the written stems: click one to open it. They're in the output folder set in Settings... ($HOME/yaas_tracks by default; --out overrides it for a single run without changing the saved setting), which Open Output Folder in the ☰ menu also opens.

Queueing jobs

You don't have to wait for a job to finish to prepare the next one: while one runs, keep browsing, and click Add to Queue (the Start button's label while a job runs) on each video you want. Queued jobs run one at a time, in order, each with the output folder and model set when it was queued.

The job list, left of the status log, shows each job's state. Right-click a job to remove it from the queue, open its output folder or show it in the browser again, or to clear the finished jobs. Double-clicking a finished job opens its output folder.

Stopping a job

Click the red Stop button to stop the running job, whatever it's doing (downloading, separating, or downloading a model); the next queued one then starts. The separation runs in a separate process, which Stop ends immediately. Stems already written are kept.

Closing Yaas while a job runs asks for confirmation, then stops it and drops the queued jobs.

Note

Respect the copyright of the video's authors. If they don't authorize sharing, keep your extracted stems for personal use only.

The menu

The ☰ button to the right of the URL bar opens a menu with:

  • Settings... — set the output folder extracted stems are written to, picked via your system's native folder picker, the separation model (see below), and the stems' sample rate: 48 kHz (default) or 44.1 kHz. The models work at 44.1 kHz; 48 kHz stems are resampled afterwards, so that DAWs working at 48 kHz can import them without converting them.
  • Open Output Folder — opens the output folder in your file manager.
  • Documentation — opens this site in your browser.
  • Report an Issue — opens the GitHub issue tracker.
  • GPU Acceleration... — on Windows/Linux, install, reinstall, or remove an on-demand CUDA-accelerated environment; see Installation. On Apple Silicon Macs, shows whether the built-in Metal (MPS) acceleration is available; see Installation.
  • About Yaas — shows the installed version and license.

Choosing a separation backend

Pick the separation model in Settings...: BS-Roformer (default), HTDemucs 6 stems, or OpenUnmix. The choice is saved for future runs. It can also be overridden for a single run from the command line.

Yaas supports two backends for splitting the soundtrack into stems:

  1. audio-separator (default) — uses the audio-separator library. Choose a model with --model:
    • roformer (default)
    • htdemucs6s
  2. OpenUnmix — uses the OpenUnmix model.
# audio-separator with a specific model
yaas --backend audio_separator --model htdemucs6s

# OpenUnmix backend
yaas --backend openunmix

Command-line options

Option Default Description
--version Print Yaas' version and exit.
-o, --out DIR Settings dialog's output folder Directory to store the output files in, for this run only.
--backend {audio_separator,openunmix} Settings dialog's model Track separation backend, for this run only.
--model {roformer,htdemucs6s} Settings dialog's model Model used by the audio_separator backend, for this run only.
--sample-rate {44100,48000} Settings dialog's sample rate (48000) Sample rate of the written stems, for this run only.