Pushing the frontiers of audio generation

Google has developed advanced speech generation technology producing natural, multi-speaker dialogue for up to 2 minutes. The system creates high-quality audio 40 times faster than real-time using a single chip, building on previous research with a specialized Transformer architecture and efficient speech codec. This technology powers features in various Google products, enhancing digital assistants and content accessibility. To prevent misuse, Google is incorporating watermarking and exploring combinations with other modalities like video.