wav2scape: New Tool for Speech Analysis Now Available
- Published
- Mon, Jul 21, 2025
- Tags
- news
- Contact
- Links
Julian Linke ´released wav2scape, a user-friendly tool that analyzes acoustic similarities and distances between speech categories using self-supervised representations. The tool leverages the pre-trained wav2vec 2.0 XLSR-53 model to extract discrete codebook representations and compute similarity matrices between different speech categories, making it applicable to languages, dialects, speaking styles and speakers. wav2scape runs entirely on local computers without requiring GPU hardware and is available as ready-to-use binaries for Windows, macOS, and Linux. The tool provides comprehensive output including similarity matrices, PCA visualizations, and detailed statistical summaries for in-depth speech pattern analysis.
You can find the GitHub repository here: GitHub Link