Preprint
SeamlessM4T: Massively Multilingual and Multimodal Machine Translation
Seamless Communication, Meta AI (arXiv), 2023 · DOI 10.48550/arXiv.2308.11596
Read the originalRead on October 16, 2026
- What it is
- The paper introducing SeamlessM4T, a single model for speech recognition and for translating speech and text across up to 100 languages, compared against cascaded systems.
- How much weight it carries
- A preprint from the model's builders. Quality figures are benchmark averages and vary by language pair.
Facts taken from this source
- The SeamlessM4T paper says that, compared with strong cascaded models, its speech to text translation into English is 1.3 BLEU points better.
“Compared to strong cascaded models, SeamlessM4T improves the quality of into-English translation by 1.3 BLEU points in speech-to-text” (Abstract)
- The SeamlessM4T authors write that conventional speech-to-speech translation systems rely on cascaded systems that perform translation progressively.
“conventional speech-to-speech translation systems rely on cascaded systems that perform translation progressively” (Abstract)
- Meta's SeamlessM4T is a single model that supports speech-to-text translation and speech recognition for up to 100 languages.
“a single model that supports speech-to-speech translation, speech-to-text translation, text-to-speech translation, text-to-text translation, and automatic speech recognition for up to 100 languages” (Abstract)
- SeamlessM4T improved into-English speech-to-text translation by 1.3 BLEU points over strong cascaded systems.
“Compared to strong cascaded models, SeamlessM4T improves the quality of into-English translation by 1.3 BLEU points in speech-to-text” (Abstract)