Whisper Large v3 Catalan Punctuated
This BSC adaptation of OpenAI Whisper Large v3 learns Catalan transcription with punctuation and capitalisation.
Model and training
Training uses 2,659 hours from Common Voice 17 and about 710 hours of AINA material. The card reports 3,369 hours and 53 minutes in total. Punctuated output comes from the fine-tuned checkpoint, rather than a display option on the plain Catalan model. BSC also publishes a faster-whisper conversion.Whisper Large v3 Catalan Punctuated: official model card
Practical use
Practical guidance. This release may reduce the separate work needed to make transcripts readable. Evaluate punctuation and word accuracy independently, especially around quoted speech and sentence boundaries. Compare short clips with longer recordings, since readable punctuation can hide a missed word or an incorrect speaker boundary.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | whisper-large-v3-ca-punctuated-3370hfaster-whisper-large-v3-ca-punctuated-3370h |
| License | Apache License 2.0.Whisper Large v3 Catalan Punctuated: 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 confirmedDownloadableThis BSC Whisper adaptation produces punctuated transcripts in Spanish, Catalan, Galician and Basque.
Launch date not confirmedDownloadable