Upload or record a short audio sample of spoken Austrian German, and this AI model will predict where in Austria the dialect most likely comes from.
Powered by a fine-tuned Wav2Vec2 model, this demo directly predicts geographic coordinates.
Due to computational limitations, please use an audio sample between 4 and 10 seconds long.
For further information, check out the paper:
Gutscher, L., Pucher, M. (2025)
Audio-Based Classification and Geographic Regression of Austrian Dialects.
Proc. Interspeech 2025, 2765-2769.
View on ResearchGate