Whisper large-v3
Suveräänsus ja majutamine
Selles juurutuses ei ole LLM EU-s majutatud: taotlustele vastab sisseehitatud mock-taustarakendus regioonis dev-mock. Ühtegi kaalu ei käitata.
Kokkuvõte
Whisper large-v3 is OpenAI’s multilingual speech-recognition and speech-translation model with 1.54B parameters under the Apache-2.0 licence. It has a fixed 30-second receptive field and processes longer audio with a sliding window, and it uses 128 mel frequency bins rather than the 80 of earlier versions. The model card warns that it was evaluated on roughly ten languages and should be evaluated in-domain before deployment.
Tehnilised andmed
- Pakkuja
- OpenAI (US)
- Konteksti aken
- Verifitseerimata: pakkuja ei ole seda väidet dokumenteerinud.
- Parameetrid
- 1.54B
- Modaalsused
- audio
- Litsents
- apache-2.0 · licence
- Avatud kaalud
- yes
- Avaldamiskuupäev
- 07. nov 2023
- Ülesanded
- translation, multilingual-support, summarization
Keeled
Tugev: en, de, fr, es, it, pt, nl, pl
Piisav: sv, da, fi, cs, ro, el, uk, ru, tr, ar, zh, ja, ko, hi
Keeled
Tugev
- en
- de
- fr
- es
- it
- pt
- nl
- pl
Piisav
- sv
- da
- fi
- cs
- ro
- el
- uk
- ru
- tr
- ar
- zh
- ja
- ko
- hi
Verifitseerimata: pakkuja ei ole seda väidet dokumenteerinud. — a language is listed as strong only where a source claims it; the card links the evidence where one exists.
Millal kasutada
Tugevused
- one model for transcription and translation to English
- long-standing tooling support across every major inference stack
- Apache-2.0 licence
Piirangud
- fixed 30-second receptive field; longer audio needs chunking logic
- the card only claims strong recognition for about ten languages
- the card advises against using it for classification or speaker attributes