Installation

Windows: using the installer

If you are on Windows and don't want to deal with Python packaging, grab the latest yaas_installer.exe from the GitHub Releases page and run it.

You also need ffmpeg installed. The simplest way on Windows is with winget (installed by default on Windows 11; available via the Microsoft Store on older versions). From PowerShell:

winget install ffmpeg

macOS: using the installer

Grab the latest yaas_installer.dmg from the GitHub Releases page, open it, and drag Yaas into /Applications.

Apple Silicon only

This build is arm64-only (Apple Silicon: M1/M2/M3/...) — GitHub retired its free Intel macOS build runners, and no free CI service fills the gap. On an Intel Mac, install via uv pip install yaas / pip install yaas instead (see below).

Unsigned application

This build isn't signed or notarized by an Apple Developer account, so Gatekeeper will warn that it's from an "unidentified developer" the first time you open it. Right-click the app and choose Open (instead of double-clicking) to bypass that warning once.

ffmpeg is included in the app: there's nothing else to install. (These are the static FFmpeg builds by Martin Riedl, licensed under the GPL and shipped as separate, unmodified programs.)

Linux: using the AppImage

Grab the latest yaas-x86_64.AppImage from the GitHub Releases page, make it executable, and run it:

chmod +x yaas-x86_64.AppImage
./yaas-x86_64.AppImage

Install ffmpeg with your distribution's package manager, e.g. on Debian/Ubuntu: sudo apt install ffmpeg.

Any platform: using uv / pip

Create and activate a virtual environment. See the uv documentation if you're not familiar with it.

Then install Yaas into that environment:

uv pip install yaas

On Windows, make sure you also have a working Python installation and ffmpeg installed (see above).

GPU acceleration

The packaged Windows and Linux installers above bundle a CPU-only build of PyTorch (a CUDA-enabled build alone exceeds GitHub Releases' 2GB per-file limit), so track separation runs on CPU by default regardless of your hardware. The macOS installer is different: see GPU acceleration on macOS below.

Installing via uv pip install yaas / pip install yaas instead pulls the regular PyPI PyTorch build. On Linux, it automatically uses a compatible CUDA GPU when one is available and falls back to CPU otherwise. On Windows, PyPI's PyTorch is CPU-only: install a CUDA build from PyTorch's own index into the same environment, for example:

uv pip install torch torchaudio torchvision torchcodec --index-url https://download.pytorch.org/whl/cu126

Enabling GPU acceleration in the Windows/Linux installers

If you installed Yaas via the Windows or Linux packaged installer and have an NVIDIA GPU with CUDA support, open the ☰ menu and choose GPU Acceleration..., then Install. This downloads a separate, self-contained Python environment with CUDA-enabled PyTorch (several GB) that Yaas then uses automatically for extraction whenever it's present. Use the same menu entry later to check its status, reinstall it (e.g. after upgrading Yaas), or remove it.

Yaas installs the CUDA 12.6 build of PyTorch, which supports the widest range of NVIDIA GPUs and drivers, or the CUDA 13 build when nvidia-smi reports a Blackwell GPU (RTX 50xx), which needs it (and an NVIDIA driver 580 or newer).

GPU acceleration on macOS

Apple hardware has no CUDA support, but on Apple Silicon Macs (M1 and later) Yaas uses the GPU through Apple's Metal (MPS) backend out of the box, with both the macOS dmg installer and pip install yaas. There's nothing extra to download. The ☰ menu's GPU Acceleration... entry shows whether Metal acceleration is available on your Mac.