aTrain
An open-source offline transcription app for Windows, Linux and macOS that uses Whisper and keeps recordings and transcripts on your device.
Alternative to:
aTrain is an offline transcription app for researchers and people working with recorded interviews. It turns speech into text on your own computer and can label segments by speaker. Recordings and transcripts stay local, so you can process sensitive material without uploading it to a transcription service.
The desktop interface uses OpenAI's Whisper through faster-whisper. It supports languages including English, German, Arabic, Chinese and Ukrainian, though transcription quality varies by language. Speaker detection uses pyannote.audio to distinguish speakers within a recording, which helps when reviewing interviews with multiple participants.
Its output fits qualitative research workflows. You can import transcripts into MAXQDA, ATLAS.ti and nVivo, where timestamps link text segments to the corresponding audio. Researchers at the University of Graz developed the tool, and researchers at Know-Center Graz tested it.
You can run aTrain on Windows, Linux and macOS. Windows and Linux have packaged desktop apps through the Microsoft Store and Flathub respectively. It can transcribe on a CPU, while NVIDIA CUDA GPUs accelerate transcription and speaker detection on Windows and Debian-based Linux. macOS processing uses the CPU and currently requires installation from the GitHub repository with the GUI extras. The optional browser/server mode is experimental.
The project is open source under AGPL-3.0. Alongside the desktop app, a command-line interface supports transcription in scripts, automation and server pipelines.
