Google Cloud launches high-definition voice model Chirp 3, supporting 248 sounds and 31 languages

Google Cloud released the high-definition voice model Chirp3 at its DeepMind headquarters in London and is open to developers through the Vertex AI platform. This model supports speech synthesis in 248 sounds and 31 languages, and is suitable for applications such as smart voice assistants and audiobooks. Chirp3 is able to capture subtle changes in human tone and make conversations more natural. In addition, Google Cloud provides a text-to-speech API, allowing users to create custom voices, but to prevent abuse, voice cloning is limited.

At the press conference, Google Cloud CEO Thomas Curian emphasized Google's vision to provide a wide range of models, including Gemini, Imagen, etc. DeepMind CEO Demis Hassabis introduced Gemini's multimodal understanding ability, such as processing YouTube video content. Google has also announced plans to improve UK AI skills and supports startups to use cloud infrastructure and AI tools.

In terms of privacy and compliance, Google promises data residency, Vertex AI and Agent Space AI tools help organizations train and serve models in compliance with local laws, especially for industries such as healthcare and finance.

Project: https://cloud.google.com/text-to-speech/docs/chirp3-hd