Whisper Catalan–Spanish Code Switching
BSC's whisper-timestamped-cs is a research adaptation for speech that switches between Catalan and Spanish.
Research release
Model and training
The checkpoint fine-tunes OpenAI Whisper Large v3 on two hours of CAESAR-TINY data. The cs suffix refers to code switching, not the Czech language. Its example uses the whisper-timestamped runtime. This is a narrow trained adaptation, and the small specialised dataset is a material limitation.Whisper Catalan–Spanish Code Switching: official model card
Practical use
Practical guidance. Evaluate words around each language change, since a recogniser can force both languages into one spelling system. Compare the adapted model with the untouched parent on the same recordings. Treat timestamp accuracy separately from word accuracy, and include language switches absent from the small adaptation corpus.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Research release |
| Access | Downloadable |
| Model identifiers | whisper-timestamped-cs |
| License | Apache License 2.0.Whisper Catalan–Spanish Code Switching: official model card |
Related models
BSC's Catalan Whisper model is a trained adaptation of OpenAI Whisper Large v3 for speech recognition.
Launch date not confirmedDownloadableWhisper Large v3 LoS is BSC's multilingual adaptation for Spanish, Catalan, Galician and Basque speech.
Launch date not confirmedDownloadable