Gemini Live Googles New Voice AI Makes Waves

Google launched Gemini Live during its Made By Google event in Mountain View, captivating attendees with its latest AI chatbot capable of semi-natural spoken conversations. TechCrunch took it for a spin.

Gemini Live is Google’s direct response to OpenAI’s Advanced Voice Mode in ChatGPT, already in a limited alpha test. While OpenAI demoed it first, Google is the first to bring a finalized version to the public.

Using Google’s latest large language model, Gemini Live shines in low-latency, verbal interactions. Answering questions in under two seconds, it feels more natural than chatting with Siri, Alexa, or even texting via ChatGPT. Interruptions don’t throw it off much either, maintaining a fairly smooth conversation flow.

You begin by choosing from 10 different voices, crafted with the help of voice actors, providing far more variety than OpenAI’s three options. The human-like quality of these voices adds a touch of realism to the conversations.

To test its capabilities, a Google product manager asked Gemini Live to recommend family-friendly wineries near Mountain View with outdoor areas and nearby playgrounds. Gemini Live suggested Cooper-Garrod Vineyards in Saratoga but made an odd recommendation about a playground called Henry Elementary School Playground, which doesn’t actually exist nearby. The nearest Henry Elementary School is more than two hours away! Clearly, there are still some quirks to iron out.

One feature Google highlighted is Gemini Live’s ability to pivot and pick up mid-sentence interruptions. While useful, this didn’t always work perfectly in practice, sometimes resulting in awkward overlaps between users and the AI.

Interestingly, users can’t make Gemini Live sing or mimic voices outside of the provided 10, likely to avoid legal issues. Also, it’s not designed to understand emotional tones in a user’s voice, unlike OpenAI’s claims during their demo.

Google acknowledges that Gemini Live is just one step on the journey to Project Astra, their vision of a fully multimodal AI model. Currently, Gemini Live is limited to voice interactions, but future updates aim to include real-time video understanding, taking the interaction to another level.

Gemini Live impresses but isn’t flawless. This new voice feature by Google feels like a significant step forward in hands-free AI interactions, yet it shows there’s room for improvement. The tech community eagerly awaits to see how Google will refine and expand this promising technology.

About The Author

Emmanuel Kesse

See author's posts