Most of the focus of the Generating AI is in a text-based interface used to generate text, images, etc. The next wave looks like a voice, rolling fast. With the latest developments, Google today announced that it will be adding Chirp 3 (speaking model from speech two text and HD text) to its vertex AI development platform starting next week.
Last week, Google quietly announced that Chirp 3 will be rolling out eight new voices in 31 languages. Platform use cases include building voice assistants, creating audio books, developing support agents, and over-vide video audio. The news was announced at an event at Google's Deepmind Offices in London.
That effort is happening at the same time as others are moving forward with their voice AI work. Last week, the startup behind Sesame – the viral, very realistic sound “Maya” and “Miles” AI apps announced the launch of models for developers to build their own customized apps and services in addition to technology.
In particular, around the Chirp 3 there are usage restrictions to try to deal with misuse. “We're working on some of these things together with our safety team,” said Thomas Kurian, CEO of Google Cloud at today's news event.
ElevenLabs is one of the leading startups that raised hundreds of millions of people to expand their work with AI Voice Services.
This news brings the Chirp 3 to the same stability as the new version of the flagship LLM, Gemini, which is being tested. It also brings image generation models and expensive Veo 2 video generation tools.
It has yet to be confirmed whether what Google is releasing on Chirp 3 is as “realistic” as some of the other AI efforts to create “human” voices (particularly Sesame's work stands out). However, as Deepmind CEO Demis Hassabis highlighted, this remains a marathon, not a sprint.
“This idea in recent times… [AI is] All the silver bullets for the next few years, I don't think it's happening yet. I think we are still a few years away from what we have with AGIs happening,” he said. “It's going to change things… over the next decade, it's mid- to long-term. That's one of those interesting moments.”
Google launched Vertex AI Way in 2021 as a platform for developers to build machine learning services in the cloud. That was, of course, well before the explosion of interest in AI, particularly generative AI, accompanied by the launch of Openai's GPT service.
Since then, the company has been leaning towards apex AI as it has been keeping up with other companies like Microsoft and Amazon. They also build generic AI tools for developers. In addition to building a generated AI on top of Gemini, developers can use Vertex AI to classify data, train models, and set up production models. It will be interesting to see if it moves to extend its walled gardens with models beyond those created by Google itself.
Google has been building a “Chirp” voice service for many years, and will return to using the name as the code name for its early efforts to compete with Amazon's Alexa service.