French AI research institute releases 'Moshi,' an AI voice assistant that can express emotions in real time
The French AI research institute
Kyutai unveils today the very first voice-enabled AI openly accessible to all
(PDF file) https://kyutai.org/cp_moshi.pdf
A video of Kyutai's presentation in Paris has been released, showing a demonstration of Moshi in action.
First, Moshi was introduced to the audience and responded, 'Hi, my name is Moshi. What can I help you with today?'
When asked about its creator, he smoothly explained about Kyotai, saying, 'It was developed by Kyotai, a non-profit AI research organization.'
When the person taking the turn said, 'I'm thinking of climbing a mountain,' the speaker responded, 'That would be wonderful,' while the speaker was still trying to continue the conversation. There was almost no sense of waiting for the dialogue to be output, and the speaker responded in real time.
When the third interlocutor asked him to 'read a poem about Paris with a French accent,' he immediately began reading the poem: 'Paris, the city of love, where time stands still and the heart takes flight.'
When asked to 'act like a pirate,' Moshi responded by becoming a pirate himself. 'What's your name? How's the pirate life? It's a wild ride, isn't it, buddy? There's freedom, danger and adventure everywhere,' he said, acting like a captain. At times, the interlocutor had a hard time stopping Moshi.
Anyone can experience the same content as this demonstration online.
moshi.chat
Kyutai developed Moshi from scratch in six months with a team of eight people. It can also be installed locally, so it can run safely on devices that are not connected to the Internet.
Through Moshi, Kyutai is demonstrating its intention to contribute to open AI research and the development of the entire ecosystem.
Related Posts:
in Web Service, Video, Posted by logc_nt