A deepfake is a video, photo, or audio recording that seems real but has been generated or manipulated with AI. The underlying technology can replace faces, manipulate facial expressions, synthesize faces, and speech. Deepfakes can depict someone appearing to say or do something that they never said or did.

No, we do not share our database for two reasons: 1. We want to prevent leaks to bad actors. 2. We want to avoid compromising the accuracy of our evaluations (if our evaluation set is used to train other detectors, it will eliminate our ability to measure their quality.)

Our technology can analyze suspicious media across audio, images, and videos. The max file size is 100MB. You can also add your own file in the following formats:
- Video: mp4, webm, avi, mkv, wmv, mov
- Image: gif, jpg, png
- Audio: mp3, wav

You can copy and paste links directly from social media posts to analyze the images, audio, and video on those posts.

Accepted direct links: TikTok, X, Mastodon, YouTube, Reddit, Facebook, Instagram, and Google Drive.

We use a set of different AI detectors, some we developed in-house, but most from technical partners. We partner with the best detection companies in the world to bring all their cutting-edge technologies to our users. Our detectors look at different aspects, which are broken down into four categories:

1. Face Manipulation - Distinguishes deepfakes from real faces or if other methods were used such as face blending, swaps, or re-enactment.

2. Generated AI - Detects if the image was created with popular tools, specifically Dall-E, Stable Cascade, Stable Diffusion XL, CQD Diffusion, Kadinsky, Wuerstchen, Titan, Midjourney, Adobe Firefly, Pixart, Glide, Imagen, Bing Image Creator, LCM, Hive, Deepfloyd, and any Generative Adversarial Network (GAN.)

3. Visual noise - Detects if artifacts from manipulation or generation are present in an image, including variation in pixels and color variation. When an AI tool creates or modifies an image sometimes types of visual noise remain.
‍
4. Audio - Detects if there are traces that audio has been manipulated or cloned.

Heavily edited videos with many quick cuts can trigger a suspicious verdict from our detectors which look for discontinuities. Short-form, multi-clip videos like those on TikTok, Instagram Reels, or YouTube Shorts can make analysis less reliable.

No. We understand that there is an important need for tools that will protect children from the dangers of deepfakes. It’s against our user terms of service to add nude or child images to our tool. Our focus is on the political landscape and elections.

No, our tool works in all languages. We can analyze non-English audio with varying results.

TrueMedia.org is a non-partisan, non-profit organization. The lead financial backer is Camp.org, and we accept contributions from individuals and other organizations. Consider a direct donation.