ElevenLabs, known for its AI voice cloning and text-to-speech models, has introduced an AI Voice Isolator. This tool allows creators to remove unwanted ambient noise from various content types, such as films, podcasts, and YouTube videos. The Voice Isolator is available for free on the ElevenLabs platform, albeit with some usage limits.

The AI Voice Isolator addresses the common issue of background noise in recorded content. During post-production, users can upload their content, and the tool processes the file to detect and remove unwanted noises, delivering clear dialogue. ElevenLabs claims the output quality is comparable to studio-recorded content. A demo showed the tool effectively removing the noise of a leaf blower to produce clear speech.

In real-world tests, the Voice Isolator successfully removed various background noises, such as door openings, table banging, clapping, and household item movements, without distorting the speech. However, it struggled with sounds like wall banging and finger snapping. The tool does not currently work on music vocals but may have some success with certain songs.+

Future Improvements and Availability

While the tool's ability to handle irregularly occurring background noise is notable, there is room for improvement. ElevenLabs has not disclosed details about the underlying models or whether user recordings are used for training. Users can opt out of data usage for training via a form in the company's privacy policy.

Currently, the Voice Isolator is available only through the ElevenLabs platform, with plans to offer API access soon. The free plan allows for 10 minutes of audio processing per month, with paid plans starting at $5/month for larger files.