Introducing Pindrop® Pulse™ Inspect, our latest innovation to combat deepfakes and prevent the spread of misinformation

Almost six months ago we started our pinhead^Ⓡ pulse^TM Solution, a cutting-edge deepfake detection technology for our enterprise customers that helps detect AI-generated voices in their call centers. Since then, we have worked with news organizations, governments, the music and entertainment industry, and corporate security teams to evaluate hundreds of suspected deepfakes. From AI-generated robocalls aim to suppress voters To sophisticated smear campaignsand from general misinformation about conflicts worldwide To Attempts to distort public perception– each case underscores the urgent need for robust deepfake detection mechanisms.

The consequences of these deepfakes are profound: they threaten the integrity of news organizations, social media platforms, and elections worldwide. The potential of misinformation to influence public opinion and disrupt social order is a harsh reality we now face.

In response to these serious threats, we are pleased to announce Pindrop^Ⓡ pulse^TM Inspect in Preview, an audio deepfake detection solution to support fact-checkers, disinformation experts, security departments, trust and safety teams, and social media platforms. As a forensics tool, Pindrop Pulse is designed to detect AI-generated speech in audio or video media, including digital media (e.g.Deepfakes in social media) and phone call media (e.g.Voicemails). Users log into the web application, upload their media files, and receive a decision within seconds on whether the content contains AI-generated speech. Additionally, users can integrate the award-winning Pindrop Pulse software. Integrate deepfake detection technology programmatically into your own workflows via our easy-to-use APIs.

Simply put, “deepfakes” are images, text, videos and audio files altered by AI.

Specifically, for speech, this means creating highly realistic audio clips that can convincingly imitate a person's voice by training an AI model on the person's publicly available speech.

This problem is growing for several reasons. First, technology has advanced to the point where the quality of synthetic speech is remarkably high. Second, commercial platforms offering these services have become incredibly affordable. And the number of tools available to create deepfakes, i.e. text-to-speech (TTS) and speech-to-speech (STS), has exploded in the last two years, so that there are now almost 2000 open source text-to-speech tools on Huggingface alone.

Humans are notoriously bad at detecting deepfakes. In one study, people were only able to detect fake audio in 54.5% of cases.and in the real world, it's even harder to distinguish between real and fake audio. Scammers who create these deepfakes are becoming increasingly sophisticated, often adding background noise or music or using very short voice clips to make detection more difficult. These scammers are constantly evolving their techniques, making it imperative for us to stay one step ahead in the fight against disinformation.

Over the past 13 years, Pindrop has built a platform based on real-time analysis of over 5 billion audio interactions. We hold over 270 patents on speech and security and 25 patents on audio deepfake detection alone. Today, we are proud to combine our experience and technology into a tool that helps combat the most deceptive audio deepfakes, especially for news media or organizations that rely on the accuracy of their content to maintain their customers' trust and their organization's credibility.

Pindrop works with some of the market and technology leaders fighting misinformation online. For example: TrueMedia.org was among the first to test our solution in their workflows and reported that Pindrop Pulse's audio deepfake detection showed higher accuracy in detecting synthetic speech than other alternatives.

According to Oren Etzioni, CEO of TrueMedia.org, “TrueMedia.org is a non-profit, non-partisan AI project to combat disinformation in political campaigns by identifying manipulated media. Our extensive evaluation found that Pindrop's audio deepfake detection has higher accuracy than other alternatives in detecting synthetic speech. We are excited to partner with Pindrop on this mission and incorporate Pindrop's deepfake detection technology into the solution for our customers and users around the world.”

Pulse Inspect provides trust and security teams with a forensic tool to improve their disinformation detection workflows.

• First-class performance: Pindrop has trained its deepfake detection model on over 370 deepfake generation tools with over 20 million statements (both real and synthetic), allowing us to achieve over 99% accuracy against previously known deepfake models and 90% of zero-day attacks using new or previously unknown tools.We have also received confirmation from third parties that our solution saves more than 40 percentage points higher accuracy than competing solutions in the audio sector.
• Resilience: News and social media are global businesses and need help detecting deepfakes in different languages. Pindrop^Ⓡ pulse^TM Inspect is language independent and the underlying training models have been tested and validated in over 40 languages, covering over 90% of the languages ​​spoken on the Internet.This technology provides resilience against hostile attacks such as adding noise, reverberation or voice alteration.
• Wide audio spectrum: The same Pindrop Pulse technology that identifies over a million call center social engineering attempts has now been extended to digital media. Pulse Inspect supports both phone call audio (8 kHz) and high-fidelity social media audio (44.1 kHz). It also provides recognition capabilities whether synthetic speech is created using text-to-speech, speech-to-speech, or voice conversion techniques.
• Video support: Pulse Inspect supports the detection of audio deepfakes in videos. The platform analyzes video files for AI-generated speech by extracting audio content from video media types.

Explainability: Pulse Inspect provides segment analysis of uploaded media to help detect partial deepfakes. This feature provides users with a visual indicator to help them determine which segment in a long-form media file was synthetically generated and which segment most likely does not contain synthetic speech.

With Pulse Inspect in preview, we invite those responsible for identifying and reporting deepfakes to test our technology for free..

Request access to a free trial here.

1. https://www.pindrop.com/blog/pindrop-named-a-winner-in-the-ftc-voice-cloning-challenge
2. https://synthical.com/article/c51439ac-a6ad-4b8d-82ed-13cf98040c7e
3. https://www.pindrop.com/blog/exposed-the-truth-about-zero-day-deepfake-attacks-metas-voicebox-case-study
4. In the NPR studyPindrop correctly identified 81 of 84 possible (96.4%) speech samples. Its closest competitor, on the other hand, identified 47 of 84 (56% – excluding samples identified as inconclusive).
5. Statista: Most commonly used languages ​​for web content (as of January 2024)
6. The general terms and conditions apply.