ElevenLabs is trying to stop the creation of misinformation with AI. For this reason, the company is making an important update to its AI speech tool.
ElevenLabs released a beta version of its AI speech tool in January. This converts text to spoken language using generative AI. According to the company, great things have resulted from the tool in recent months, such as an audiobook and a radio channel working entirely with AI.
First step to a solution
In turn, other users experimented with the ability to create audio deep fakes. They use the tool to make an audio recording in which it seems a celebrity or friend says certain things.
The company now wants to solve that with AI Speech Classifier. It does not stop the creation of deceptive speech messages but allows users to check the authenticity of a speech message.
However, the tool can only check if the audio file was created using ElevenLabs. If the AI speech tool’s original audio file was tinkered with, AI Speech Classifier also becomes more inaccurate. The reliability then drops from 99 percent to about 90 percent.
“The AI Speech Classifier is a critical step forward in our mission to develop efficient tracking for AI-generated media. With today’s launch, we seek to further reinforce our commitment to transparency in the generative media space,” the announcement reads.