Former co-developer of voice mode for OpenAI’s ChatGPT launches WaveForms AI to make AI voice interactions feel more intuitive
One of the lead researchers behind the voice capabilities for OpenAI’s ChatGPT has raised $40 million (£31m) from venture firm Andreessen Horowitz for an artificial intelligence start-up that aims to give a more natural feel to conversations between humans and computers.
Alexis Conneau, the co-founder and chief executive of start-up WaveForms AI, was a lead developer of the voice mode of OpenAI’s GPT-4o model.
The mode featured the ability to respond to users without a delay and to handle interruptions, both capabilities being seen as significant advances to give AI voice interactions a more intuitive feel.
The venture funding valued WaveForms AI at $200m, the company said.
Emotive engagement
Conneau left OpenAI in August and began working on WaveForms AI in September, the same month that the voice feature rolled out to all ChatGPT paid users.
He said the company aims to make AI models that can pick up on voice cues and take account of them in its interactions.
Hesitations, for instance, could lead the model to be more reassuring, he said.
The company as yet has no product and only five employees, including co-founder Coralie Lemaitre, who is married to Conneau and previously worked at Google.
After the launch of the ChatGPT voice feature, OpenAI chief executive Sam Altman posted the word “her” on social media platform X, which was seen as a reference to the 2013 film of the same name in which a man falls in love with an AI assistant voiced by Scarlett Johansson.
Voice interaction
Days later Johansson criticised the company for the voice “Sky”, which she said sounded “so eerily similar” to her own, after she had declined to allow her own voice to be used.
OpenAI said it did not intend the voice to be similar to Johansson’s, but removed it.
OpenAI is one of several companies seeking to make more realistic and engaging voice assistants and customer service agents.
