Widely-used open multilingual speech recognition and translation model.
Specifications
- Provider
- OpenAI
- Type
- Open-source / open-weight
- Modality
- Speech-to-text
- Category
- Speech model
- License
- MIT
- Released
- September 21, 2022
What it was trained for
An automatic speech recognition model trained on a large multilingual, multitask audio dataset to transcribe and translate spoken audio.
Best for
- ▸Speech-to-text transcription
- ▸Multilingual audio transcription
- ▸Translation of speech into English
- ▸Subtitle and caption generation
- ▸Voice interface and meeting-note pipelines
Capabilities
Audio inputMultilingual transcriptionSpeech translationOpen weightsSelf-hostable
Performance & positioning
Robust across accents, background noise, and many languages; a widely adopted baseline for open speech recognition.
More from OpenAI