Description of the tool
aTrain DU Edition is a tool for automatic transcription of audio and video files. The program uses modern AI models and operates entirely offline, ensuring that no audio files are sent to external servers. The tool is based on open-source code and supports various transcription models, including KB-Whisper.
The university’s IT department has adapted the original code to make the tool adapted for use within HDa. The tool is provided as-is, and unfortunately, we cannot offer qualified guidance or extensive support. aTrain DU Edition can be downloaded and used free of charge.
Getting started
-
Install aTrain DU Edition
Log in to Företagsportalen, where Windows users can install aTrain DU Edition. The program is installed locally, and transcription takes place directly on your computer. -
Download models
In aTrain DU Edition, you can work with several AI models, which are downloaded via the Model Manager. One of these is KB-Whisper, a speech-to-text model developed by the National Library of Sweden. Learn more about the model on the KB website: Vältränad AI-modell förvandlar tal till text | Kungliga biblioteket – Sveriges nationalbibliotek -
Start transcription
Select the audio file, model, and language.
If the file contains multiple speakers, enable Multispeaker. You can specify the number of speakers yourself or let the tool detect up to 10 speakers automatically. This feature works but is still limited, and results with Multispeaker may vary. -
Retrieve transcribed files
Transcriptions are saved locally on your computer.
When the transcription is complete, click Open in the program to view your files. Files are stored at “C:\Users\username\aTrain\transcriptions”.
Several export formats are available.
Citing aTrain in research publications
aTrain was developed by researchers at the Business Analytics and Data Science Center at the University of Graz. If you use aTrain in publications or research projects, you should cite the following article:
Haberl, A., Fleiß, J., Kowald, D. & Thalmann, S. (2024). Take the aTrain. Introducing an Interface for the Accessible Transcription of Interviews. Journal of Behavioral and Experimental Finance, 41, 100891.
Storage of research data
As a researcher, you are responsible for moving the generated text material from the location where it is automatically stored during transcription (on your computer) to Dalarna University’s central storage space with backup, in accordance with the Rules for the Storage of Digital Information (pdf).