Pindrop claims to detect AI audio deepfakes with 99% accuracy

Today, Pindropa company that provides voice security, identity verification and fraud detection solutions, has announced the release of Pulse Inspect, a web-based tool for detecting AI Generated Speech in any digital audio or video file with a claimed very high degree of accuracy: 99%.

The feature is available in preview as part of Pindrop’s Pulse product suite and offers detection regardless of the tool or AI model that generated the audio.

This is a striking and ambitious offering from common industry practice, where AI vendors only release AI classifiers to detect synthetic content generated by their tools.

Pindrop offers Pulse Inspect as an annual subscription for organizations looking to combat the risk of audio deepfakes at scale. However, CEO Vijay Balasubramaniyan tells VentureBeat that they may also launch affordable pricing tiers—with a limited number of media checks—for consumers.

“Our pricing is designed for organizations with a recurring need for deepfake detection. However, based on future market demand, we may consider launching pricing options in the future that are better suited for occasional users,” he said.

Pindrop targets the rise of audio deepfakes

While deepfakes have been around for a long time, the rise of text-based generative AI systems has made them more popular on the internet. Popular generative AI tools, such as those from Microsoft and ElfLabsare used to record the audio and video of celebritiesbusiness people and politicians to spread widespread misinformation/scams, which damages their public image.

According to Pindrop's internal reportMore than 12 million American adults know someone who has had deepfakes created without their consent. These duplicates can be anything from images to videos to audio, but they all have one thing in common: They thrive on virality, spreading like wildfire on social media.

To address this evolving problem, Pindrop announced the Pulse product suite earlier this year. The first offering in the portfolio helped enterprises detect deepfake calls coming into their call centers. Now, with Pulse Inspect, the company is going beyond calls to help organizations inspect any audio/video file for AI-generated synthetic artifacts.

Upload questionable audio files for analysis

Essentially, the offering consists of a web application where a business user can upload the suspicious file for analysis.

Previously, the entire process of checking for synthetic artifacts in existing media files required time-consuming forensic investigation. In this case, however, the tool processes the audio in a matter of seconds and comes up with a “deepfake score,” complete with sections that include AI-generated speech.

This rapid response allows organizations to take proactive measures to prevent the spread of misinformation and maintain their brand credibility.

Training and analysis process

Pindrop says it trained its proprietary deepfake detection model on more than 350 deepfake generation tools, 20 million unique utterances, and more than 40 languages. This results in a 99% detection rate of deepfake audio, based on the company's internal analysis of a dataset of approximately 200,000 samples.

The model checks media files for synthetic artifacts every four seconds, ensuring deepfakes are accurately classified, especially in cases of mixed media with both AI-generated and authentic elements.

“Pindrop’s technology leverages recent breakthroughs in deep neural networks (DNN) and advanced spectrotemporal analysis to identify synthetic artifacts using multiple approaches,” Balasubramaniyan explains.

No vendor specific detection limits

Because Pindrop has trained its detection model on tools from over hundreds of generations, Pulse Inspect has no tool-specific limitations for detection.

“There are over 350 deepfake generator systems, with many prolific audio deepfakes on social media likely coming from open-source tools rather than commercial tools like ElevenLabs. Customers need comprehensive tools like Pindrop’s that aren’t limited to detecting deepfakes from a single system, but can identify synthetic audio across all generation systems,” Balasubramaniyan added.

However, it is important to note that there may be cases where the tool cannot identify deepfakes, especially when the file contains less than two seconds of net speech or very high levels of background noise. The CEO said that the company is continuously working to address these gaps and further improve detection accuracy.

Currently, Pindrop Pulse Inspect is targeting organizations such as media companies, nonprofits, government agencies, celebrity management firms, law firms and social media networks. Balasubramaniyan didn’t share the exact number of customers using the tool, but he did say that “a number of partners” are using the product by paying for a volume-based annual subscription. This includes TrueMedia.org, a free-to-use product that helps critical election audiences detect deepfakes.

In addition to the web app supporting manual uploads, Pulse Inspect can also be integrated into custom forensic workflows via an API. This can power bulk use cases, such as a social media network flagging and removing malicious AI-generated videos.

Balasubramaniyan said the company plans to strengthen the Pulse suite by improving the explainability aspect of the tools — with a feature to trace the source of deepfake generations — and supporting more modalities.