DeeptectDeepfakeDetection
Check suspicious audio, images, or video for signs of AI generation or manipulation. Built on university research, made for everyone.
Take detection challenges, explore viral deepfakes, and stay informed and protected.
The fading line between reality and AI
Synthetic media is now good enough that looking closely is no longer enough.
Generative models improve constantly, and the visual tells people rely on are the first things newer models fix. Seen briefly or out of context, even a careful eye is easy to fool.
- Zoom into eyes, teeth, fingers, and hair, where models may still slip.
- Check that shadows and reflections match the rest of the scene.
- Trace the source before you trust or forward it.
More checks in our practical guide to spotting deepfakes.
Deepfakes are spreading worldwide, and their misuse is rising.
What once needed specialist skills and costly hardware now takes a free app. Convincing fakes are made and shared faster than ever, on platforms tuned to amplify whatever spreads fastest, not whatever is accurate.
One form of deepfake imitates individuals without their knowledge or consent, enabling impersonation, harassment, and other crimes.
Deepfakes also target large audiences with misinformation, eroding trust in digital content overall.
What we detect
Check pictures for signs of AI generation before you pass them on, and help keep misinformation from spreading.
Safeguard video conferencing and strengthen identity checks. We extract key frames and analyze those for signs of AI generation. The audio track is not yet included.
Counter voice-clone scams and undisclosed AI music. We check recordings, helping identify AI-generated or manipulated audio.
Two ways to submit
Paste a link to any publicly accessible media file. Supports and more.
Upload audio, image, or video files directly from your device.
How we detect deepfakes
We examine digital media to identify fingerprints, meaning subtle, recurring patterns that appear when content is generated or altered by AI.
Just as a camera's optical lens or sensor chip leaves tiny, characteristic traces in every photo it captures, generative models leave identifiable patterns in the output they produce.
Our detectors can analyze content far more thoroughly than a human could. They extract and evaluate a large number of carefully engineered features across different representations and scales, and combine them to reach a prediction. This lets them surface inconsistencies and patterns invisible to the human eye or ear.
How it started
This project emerged from university research in advanced deepfake detection. Research has made significant progress, and highly capable detection models already exist, yet these technologies rarely reach the people and organizations who need them in everyday digital life. We are closing that gap.
Detection is only one part of the solution. We also raise awareness of emerging deepfake trends, strengthen people's ability to recognize manipulated media through education and hands-on challenges, and enable professional use of our detectors through APIs.
The journey ahead
- NowImage deepfake detection, including frames from videos.
- NextAudio deepfake detection for voices and music.
- LaterSocial media detection bots, and an API for organizations and platforms.