Generative speech firm ElevenLabs raises $19M, launches tool to snuff out deepfakes

As the risks presented by generative AI fraud may be growing worldwide, ElevenLabs has unveiled a new tool that will allow users to upload any audio sample to identify if it contains audio generated with the company’s AI.

ElevenLabs is a U.S.-based start-up which offers a speech synthesis platform publishers and creators can use to make synthetic audio content. For example, the technology could read an online newspaper article or book in a reasonably human-sounding voice.

According to the company, its system has certain “specific, detectable characteristics.”

When users upload a sample to the AI Speech Classifier, its algorithm will supposedly scan for these characteristics.

ElevenLabs says it will then be able to confirm whether the content was generated by its platform,

The company claims it can achieve accuracy of up to 99 per cent if the input was unmodified, which falls to just 90 per cent if it underwent codec or reverb transformations before being fed into the tool.

This accuracy drops the more the content has been post-processed as per the firm. If additional audio tracks have been added, this will also affect the result.

The launch coincides with ElevenLabs raising $19 million in Series A round co-led by Nat Friedman, Daniel Gross, and Andreessen Horowitz, with further participation from Credo Ventures and Concept Ventures.

The round also included an array of angel investors such as Instagram co-founder Mike Krieger, Oculus VR Co-founder Brendan Iribe, Ubiquity6 Co-founder Anjney Midha, and Deepmind and Inflection Co-founder Mustafa Suleyman, according to a company blog post.

ElevenLabs says it will use the new investment to build its research hub for voice AI as well as to launch additional products to support market verticals such as publishing, gaming, and entertainment.

The potential of deepfakes for fraudulent use is an issue which is attracting much interest worldwide.

Instances of deepfake fraud doubled in North America between 2022 and Q1 2023 according to a new report from London-based identity verification platform Sumsub.

The proportion of these type of incidents compared to total fraud cases, jumped from 0.2 per cent to 2.6 per cent in the U.S, and from 0.1 per cent to 4.6 per cent in Canada.

Article Topics

AI  |  Andreessen Horowitz  |  deepfake detection  |  ElevenLabs  |  fraud prevention  |  funding  |  generative AI  |  synthetic voice