NordVPN introduces AI voice detector extension for Chrome browser

What is the AI Voice Detector?

The AI Voice Detector is a free Chrome extension that analyzes the audio content in the active browser tab to tell whether the voice is human or AI-generated. Designed by the NordLabs team (the minds behind NordVPN’s link, file, and scam text checkers), it reviews the audio in real time, without interrupting users' browsing. The feature examines acoustic characteristics only, meaning it doesn’t track, understand, record, or interpret the content of any conversation. You can add it to your Chrome browser for free through NordLabs site or on Chrome Web Store. 

It’s also worth noting that the AI Voice Detector: 

• Is based entirely in the user's Chrome browser. The extension uses Chrome's built-in extension APIs and runs within the browser's sandboxed environment.
• Doesn’t use server components. There is no backend, no cloud service, and no API endpoint involved in detection.
• Doesn’t require network calls for inference. The machine learning (ML) model and all processing code are bundled with the extension and executed locally. The only network requests the extension makes are lightweight analytics events.
• Doesn’t send audio outside the device. Captured audio is processed in-memory, used only for real-time feature extraction and classification, and discarded when detection is stopped or the tab is closed.

How does the AI Voice Detector work?

The AI Voice Detector works by capturing the audio stream from the active browser tab and running it through an on-device neural network model in real time. The whole process takes place without any interruption, so users don’t have to pause or wait until the analysis is done. Results are displayed continuously through the extension popup and as a small, unobtrusive notification on the webpage itself. These notifications are color-coded for clarity: Green for human speech, red for AI-generated, and amber for potential AI. 

Below you can find a more detailed description of how AI Voice Detector works:

1. 1.Audio capture. When the user activates detection, the extension captures the audio stream from the current browser tab using Chrome's built-in tab capture API. The audio continues to play normally for the user.
2. 2.Feature extraction. The captured audio is resampled to a standard rate (16 kHz) and broken into short overlapping segments (2 seconds each, with 0.5-second overlap). For each segment, the extension extracts a compact representation of the audio's frequency characteristics over time — a “mel spectrogram” — along with its rate of change (“delta features”). These representations capture the acoustic fingerprint of the audio without retaining the original speech content.
3. 3.Voice activity detection. Before classification, each segment is screened for speech content. An energy-based filter discards silent segments, and a neural voice activity detector (Silero VAD) filters out non-speech audio such as music or background noise. Only segments containing speech are passed to the classifier.
4. 4.On-device classification. Each segment's features are fed into a lightweight neural network (~1,075,000 parameters: A convolutional neural network with depthwise separable convolutions, channel attention, and multi-head self-attention) that outputs a single score between 0 and 1, indicating the probability that the audio is AI-generated.
5. 5.Score aggregation. Individual segment scores are collected over a rolling window of approximately 60 seconds. The extension counts how many segments exceed the AI-detection threshold and produces an overall verdict with a confidence percentage.
6. 6.Verdict display. The verdict is shown to the user in the popup and as a color-coded overlay on the webpage (green for human, red for AI, amber for mixed).

The entire pipeline — from audio capture to verdict — runs locally in the browser. There are no cloud API calls, no server round-trips, and no external processing. When the user stops detection or closes the tab, all audio buffers are immediately discarded. The feature also doesn’t record or access user identity, browsing history, cookies, or account information whether enabled or not. 

How does the AI Voice Detector benefit your online safety? 

NordVPN’s AI voice detector extension is the first step in recognizing and combating advanced scammer threats. AI-generated voice technology is rapidly improving and becoming widely accessible. Tools capable of cloning a voice from just a few seconds of audio are within arm’s reach, making synthetic voices an increasingly common weapon in scam calls, impersonation attacks, and manipulated video content.

While an AI Voice Detector is not an ultimate solution, it’s an additional tool that can provide more comfort and security when browsing online. You won’t have to worry about whether you’re listening to an actual human voice or a carefully generated voiceover — the AI Voice Detector will check that for you, without interrupting your browsing. Paired with features such as NordVPN’s Threat Protection Pro™, this extension is a convenient way to strengthen your online security, while saving performance and time.