Google Cloud unveiled its high-definition speech model, Chirp3, at an event held at DeepMind's London headquarters. The model is now available to developers through the Vertex AI unified machine learning platform, providing a rich set of development tools to foster innovation.
Chirp3 supports 248 distinct voices and text-to-speech in 31 languages. Developers can leverage this model to create a variety of applications, including smart voice assistants, audiobooks, and video dubbing. Google claims Chirp3's speech capabilities capture the subtle nuances of human intonation, resulting in more engaging and lifelike conversations.
In addition to using pre-built voices, users can create custom voices via Google Cloud's text-to-speech API. However, to ensure responsible use and prevent potential misuse, Google has restricted access to this voice cloning feature to align with ethical AI practices.
At the launch event, Google Cloud CEO Thomas Kurian emphasized Google's overall vision of providing a broad range of models, including Chirp3, Gemini, Imagen, Veil, and others. Google also introduced Agent Space, a new product designed for enterprise users to meet their specific needs.
DeepMind CEO Sir Demis Hassabis highlighted the evolution of Gemini, particularly its multimodal understanding capabilities. He demonstrated how users can paste a YouTube link into AI Studio, and Gemini will process the video content, leveraging its long-context window to allow users to ask questions and quickly locate key moments in lectures or sporting events.
Furthermore, Google announced an initiative to boost AI skills in the UK through comprehensive training programs, empowering professionals with effective AI expertise. Google will provide UK startups with credits for cloud infrastructure and AI tools to accelerate the development and scaling of innovative solutions, stimulating entrepreneurial activity.
Regarding privacy and compliance, Google reiterated its commitment to data residency, emphasizing that its Vertex AI and Agent Space AI tools help organizations train and serve models in compliance with local regulations. This is crucial for industries like healthcare and finance, which have stringent privacy and compliance requirements.
Project: https://cloud.google.com/text-to-speech/docs/chirp3-hd
Key Highlights:
🌟 Google Cloud launches Chirp3, a speech model supporting 248 voices and 31 languages, empowering developers to build intelligent applications.
🔒 Google restricts access to voice cloning capabilities to ensure ethical AI practices and prevent misuse.
💼 Google initiates a program to enhance AI skills in the UK and provides cloud infrastructure support to startups, fostering innovation.