Very cool audio LLM demo by @kyutai_labs today at @Ircam 😎 for me the best part was their willingness to discuss their architecture and share their (6 month 13 ppl 🤯) research path towards today’s demonstration in detail! can’t wait for the code to be out 👾 #opensource https://t.co/ReIyIQVyMW
This demo by @kyutai_labs is insane 🤯 A super responsive voice AI (instant!) with the ability to vary its voice similar to GPT-4o. And it can make poems with a French accent 👨🏻🎨 It can whisper poems (really well) 🤫 And role play 👩🏽🚀 https://t.co/uNGjzyjn4b https://t.co/qv1DK1ZwrS
You can check out that new GPT-4o-Voice alternative here! Free to use, even via web! https://t.co/kVAalL0H3S

Kyutai Labs, a company with former members from llama, musicLM, and DEIT, has unveiled an open GPT-O on device. They released a video showcasing their GPT-4o-Voice competitor, receiving positive feedback for its voice model and capabilities, including instant responsiveness, voice variation, French accent poems, whispering poems, and role-playing. The demo was considered a significant advancement in audio models, with a detailed discussion on their research path and architecture.


