Kyutai, an AI nonprofit group formed last year, revealed its Moshi service at in event in Paris on Wednesday. Scientists for the lab said their system can speak with 70 different emotions and styles. They demoed the assistant offering advice on climbing Mt. Everest and reciting a poem it had composed with a thick French accent.

“It thinks as it speaks,” Kyutai Chief Executive Officer Patrick Pérez said. “We believe Moshi has tremendous potential to change the way we communicate with and through machines.”

The assistant is the latest challenger to OpenAI’s ChatGPT, the best-known chatbot. A growing number of startups and Big Tech companies, including Anthropic, Cohere and Alphabet Inc.’s Google, have rushed to introduce models to compete with OpenAI’s GPT-4, even as some industry experts worry about the dangers presented by the emerging technology. 

OpenAI in May held a launch event for a voice assistant for ChatGPT Plus users that for the first time paired powerful image-recognition capabilities with lightning-quick responses. The new product was supposed to be available within weeks, but the company later delayed the roll out until fall and said it won’t initially include video and screen-sharing features it had demonstrated. 

OpenAI also faced backlash for showcasing an AI voice in its feature that sounded like the actress Scarlett Johansson. The company pulled the voice after the actress hired lawyers. 

Pérez said his lab will release the models and research behind the assistant as open-source technology, where the code is shared freely. He called Moshi the “first real-time voice AI assistant released.”

The new service is an “experimental prototype,” Kyutai said in a statement on Wednesday. A representative for the lab said the model and research will be available in the coming weeks, without specifying a date.

Kyutai was launched in November with €300 million ($324 million) in funding, including from Niel, French billionaire Rodolphe Saadé and former Google chairman Eric Schmidt. Pérez, a former director for Valeo SA, hired researchers from Google DeepMind and Meta Platforms Inc. for his lab.

The voice assistant is a promising indicator that Europe can be a global player in AI development, according to Niel. “All the products they showed today are the best in class worldwide,” he said in an interview Wednesday. “We are very happy to have this in Europe.”

Kyutai’s Chief Science Officer Hervé Jégou addressed safety concerns at the event briefly. The lab will use indexing and watermarking tools to identify and track audio that’s generated by its AI, he said. 

To train its new model, Kyutai said it worked with a voice actress it called Alice, without disclosing her full name.