Meta has developed a new speech-to-text model, SeamlessM4T, capable of translating nearly 100 languages. The model, which can also recognize when a speaker switches between languages, represents a significant breakthrough in translation technology. It is part of Meta’s ongoing efforts to create a universal language translator. The company has also released the metadata for its open translation dataset, SeamlessAlign. The new model and dataset are expected to significantly improve language moderation for companies like Meta.