New transcription model
Pinecast now uses a new transcription AI model, from the same vendor.
The new model brings some great improvements:
- For speaker labels (noting who is talking), the new model is almost twice as good on a variety of benchmarks.
- The timing of words in your transcripts are now slightly more accurate.
- You’ll get more accurate transcription of numbers, showing
12:50instead oftwelve fiftywhen talking about time. - The accuracy of transcription is broadly improved across all of the supported langauges.
- 14 new languages are now supported, including Urdu, Russian, Korean, Catalan, Galician, Romanian, Estonian, Persian, Cantonese, Afrikaans, Marathi, Zulu, Xhosa, and Norwegian Nynorsk.