Ircam Amplify Identifies AI Voice Cloning

Ircam Amplify, a provider of artificial intelligence audio technology, has developed AI Speech Detector to help address the accelerating threat of voice deepfakes.

Ircam Amplify's AI Speech Detector enables the swift identification of artificial speech before it reaches audiences. It detects cloned voices in audio files and verifies audio integrity with 98 percent accuracy without the need for a reference voice for comparison. The solution integrates into existing workflows and technology infrastructures. It can analyze up to 20 audio files per API call and process hundreds of calls simultaneously. For each file, it determines whether AI was used in its creation and provides a confidence score, indicating the percentage of the file that has been generated by AI.

Ircam Amplify has trained the solution on 10 voice genAI models, including the top five commercial models and the top four most used open-source models. As an agnostic solution, it can add new models in less than three weeks.

"As generative AI becomes increasingly accessible, so too does the risk of its misuse. Voice deepfakes pose a significant and growing threat to media integrity, brand reputation, and public trust. Our AI Speech Detector delivers the transparency and security that media outlets, fake news detection and social platforms need in an era where misinformation can spread around the world at an accelerating pace," said Nathalie Birocheau, CEO of Ircam Amplify, in a statement.

"Using the very latest technology developed in our audio research lab, the AI Speech Detector is a powerful and scalable solution that doesn't require a reference voice for comparison. We have taken a close look at the market needs and the key actors in the sector to ensure that we are targeting popular and accessible models that can be the source of the most scurrilous activity," said Romain Simiand, chief product officer of Ircam Amplify, in a statement.