Meta has introduced an AI translation model named SeamlessM4T, which supports speech-to-text, text-to-text, and text-to-speech translations for nearly 100 languages. The model's speech recognition capabilities have reached human levels, and its ability to handle background noise and varying speech has significantly improved. Meta has made SeamlessM4T available for free under a research license and has also released the key training dataset, SeamlessAlign, which is the largest publicly available dataset for multimodal translation to date. The launch of SeamlessM4T is considered a major step towards a world without language barriers. Although the model still has some biases and errors, Meta plans to conduct further research and improvements based on SeamlessM4T.