Whisper large-v3

model·speech recognition model·active

Multilingual ASR and speech-to-English translation checkpoint.

Verified facts

Official site	https://huggingface.co/openai/whisper-large-v3 ↗
Geography	unknown
parameter count	1.55B
huggingface repo	openai/whisper-large-v3
api	unknown
inputs	audio
license	MIT
outputs	text transcription
weights	public
modalities	audio; text transcription
dependencies	unknown
model family	Whisper
access method	downloadable weights; openai-whisper Python package; Hugging Face Transformers
exact version	large-v3
repository links	https://huggingface.co/openai/whisper-large-v3; https://github.com/openai/whisper
evaluation results	result: OpenAI reports 10% to 20% error reduction versus large-v2 for evaluated languages below 60% error. · source: https://github.com/openai/whisper/discussions/1762 · benchmark: Common Voice 15 and Fleurs
evidence separation	developer claims: · measured results: · independent evaluations:
developer maintainer	OpenAI
material us relevance	Publicly accessible in or materially offered to the United States market as of snapshot.
documented capabilities	multilingual automatic speech recognition; lyric/vocal transcription as a downstream musical workflow
products exposing model	unknown
commercial use conditions	MIT terms apply to code and weights.
known geographic or policy restrictions	unknown
safety watermarking disclosure provenance	unknown