Meta has launched the Omnilingual ASR, an open-source automatic speech recognition system capable of understanding over 1,600 languages, including 500 low-resource languages previously unsupported. "This is a major step towards truly universal AI," said Alexandr Wang, Meta's AI chief. The system employs a seven billion parameter wav2vec 2.0 model trained on diverse datasets, aiming to enhance accessibility in digital speech technology and bridge the gap for underrepresented linguistic communities.
Meta has launched the Omnilingual ASR, an open-source automatic speech recognition system capable of understanding over 1,600 languages, including 500 low-resource languages previously unsupported. "This is a major step towards truly universal AI," said Alexandr Wang, Meta's AI chief. The system employs a seven billion parameter wav2vec 2.0 model trained on diverse datasets, aiming to enhance accessibility in digital speech technology and bridge the gap for underrepresented linguistic communities.