Whisper BSC Large v3 Catalan
BSC's Catalan Whisper model is a trained adaptation of OpenAI Whisper Large v3 for speech recognition.
Model and training
BSC fine-tuned the parent model on about 4,700 hours of Catalan recordings from Project AINA. It produces transcripts without punctuation. The official faster-whisper repository converts this checkpoint for another inference runtime; it is not a new recognition model. OpenAI supplies the foundation architecture and weights, while BSC supplies the Catalan adaptation.Whisper BSC Large v3 Catalan: official model card
Practical use
Practical guidance. Use this release when the next processing step can handle plain transcripts. Assess accents and names in recordings that resemble the intended source. Keep punctuation restoration separate from recognition evaluation. Compare the runtime conversion on speed and output consistency without counting it as an independent trained model.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | whisper-bsc-large-v3-catfaster-whisper-bsc-large-v3-cat |
| License | Apache License 2.0.Whisper BSC Large v3 Catalan: official model card |
Related models
This BSC adaptation of OpenAI Whisper Large v3 learns Catalan transcription with punctuation and capitalisation.
Launch date not confirmedDownloadable- Whisper Valencian
Barcelona Supercomputing Center · BSC Whisper
BSC's Valencian Whisper release adapts OpenAI Whisper Large v3 to Valencian-accented Catalan speech.
Launch date not confirmedDownloadable