Whisper large-v3
Suverenita a hostování
V tomto nasazení není hostováno na LLM EU: požadavky odpovídá vestavěný mock backend v regionu dev-mock. Neběží žádné váhy.
Souhrn
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.
Specifikace
- Poskytovatel
- OpenAI (US)
- Kontextové okno
- Neověřeno: poskytovatel tento údaj nedokumentoval.
- Parametry
- 1.54B
- Modality
- audio
- Licence
- apache-2.0 · licence
- Otevřené váhy
- yes
- Datum vydání
- 07. 11. 2023
- Úlohy
- translation, multilingual-support, summarization
Jazyky
Silná: en, de, fr, es, it, pt, nl, pl
Dostatečná: sv, da, fi, cs, ro, el, uk, ru, tr, ar, zh, ja, ko, hi
Jazyky
Silná
- en
- de
- fr
- es
- it
- pt
- nl
- pl
Dostatečná
- sv
- da
- fi
- cs
- ro
- el
- uk
- ru
- tr
- ar
- zh
- ja
- ko
- hi
Neověřeno: poskytovatel tento údaj nedokumentoval. — a language is listed as strong only where a source claims it; the card links the evidence where one exists.
Kdy jej použít
Silné stránky
- one model for transcription and translation to English
- long-standing tooling support across every major inference stack
- Apache-2.0 licence
Omezení
- 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