Whisper large-v3

OpenAI

Netalpinamas — imitavimo vidinis serveris Atvirieji svoriai

Suverenitetas ir prieglobstis

Netalpinamas — imitavimo vidinis serveris

Šiame diegime netalpinamas LLM EU: į užklausas atsako įtaisytasis imitavimo vidinis serveris regione dev-mock. Jokie svoriai nepaleidžiami.

Santrauka

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.

Specifikacijos

Teikėjas
OpenAI (US)
Konteksto langas
Nepatikrinta: teikėjas nedokumentavo šio teiginio.
Parametrai
1.54B
Modalumai
audio
Licencija
apache-2.0 · licence
Atvirieji svoriai
yes
Išleidimo data
2023-11-07
Užduotys
translation, multilingual-support, summarization

Kalbos

Stipri: en, de, fr, es, it, pt, nl, pl

Pakankama: sv, da, fi, cs, ro, el, uk, ru, tr, ar, zh, ja, ko, hi

Kalbos

Stipri

  • en
  • de
  • fr
  • es
  • it
  • pt
  • nl
  • pl

Pakankama

  • sv
  • da
  • fi
  • cs
  • ro
  • el
  • uk
  • ru
  • tr
  • ar
  • zh
  • ja
  • ko
  • hi

Nepatikrinta: teikėjas nedokumentavo šio teiginio. — a language is listed as strong only where a source claims it; the card links the evidence where one exists.

Kada naudoti

Stiprybės

  • one model for transcription and translation to English
  • long-standing tooling support across every major inference stack
  • Apache-2.0 licence

Apribojimai

  • 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