installation.md

Installation (end users)

aTrain is published as a packaged desktop app and can also be installed from source with pip.

Setting up aTrain for development? See CONTRIBUTING.md, which uses the uv workflow. Building a standalone executable is also covered there.

Packaged apps (recommended)

• Windows: Microsoft Store
• Linux: Flathub (see also the Linux installation guide for a manual command-line setup)

Additional download types are listed on the university download page.

Rolling aTrain out to managed machines? See Windows deployment.

Removing aTrain again: see Uninstalling aTrain.

How release builds are signed: see the Code signing policy. How to check a downloaded release: see Verifying a release.

Install from source with pip

You need Python ≥ 3.11.

Set up and activate a virtual environment:

python -m venv venv
# Windows
.\venv\Scripts\activate
# macOS / Linux
source venv/bin/activate

Install aTrain with the GUI extras (needed for aTrain start, the desktop / browser app):

pip install "aTrain[gui] @ git+https://github.com/aTrainTranscription/aTrain.git"

On Windows, prepend the PyTorch CUDA index so pip pulls the CUDA torch wheel:

pip install "aTrain[gui] @ git+https://github.com/aTrainTranscription/aTrain.git" \
    --extra-index-url https://download.pytorch.org/whl/cu128

On Linux the PyPI torch wheel already bundles CUDA. macOS is CPU-only. NVIDIA CUDA GPU support currently covers Windows and Debian-based Linux.

Download the models for transcription and speaker detection. This only has to be done once:

aTrain init

Start aTrain:

aTrain start

Command-line / headless usage

For headless transcription pipelines, aTrain also exposes a CLI (aTrain_core transcribe). See the "Headless / CLI Usage" section in the README.