SeamlessM4T
One model for speech recognition, speech/text translation and text-to-speech across ~100 languages, CC BY-NC.
Unified five translation/transcription tasks in a single multitask model with ~100 input languages, trained with the 470,000-hour SeamlessAlign mined dataset.
- Date
- Tuesday, 22 August 2023
- Lab
- Meta
- Kind
- open-weights
- Access
- open weights (restricted license)
Figures
| Measure | Value | Measured by |
|---|---|---|
| Language coverage | ~100 languages (35 + English speech output) | company |
Non-commercial license.
Sources
This record was checked against its sources on 6 October 2026. How we check