SeamlessM4T
Paper preview
Paper summary
Meta's SeamlessM4T is a unified multilingual and multimodal machine-translation system that handles five translation tasks in one model.
Ask this paper
01
Five tasks, one model: Handles ASR, text-to-text, speech-to-text, text-to-speech, and speech-to-speech translation in a unified architecture.
02
100+ languages: Covers ~100 languages for text and ~36 for speech, dramatically broadening the set of supported language pairs compared to prior systems.
03
Unified training: Avoids the cascade of per-task models typical in translation pipelines, reducing error accumulation and improving multilingual generalization.
04
Open release: Releases model weights and evaluation code, providing a strong open baseline for multilingual multimodal translation research.