# Authini

> Authini (authini.co) is an AI video detector. It checks whether a video looks like real camera footage or was generated by OpenAI's Sora 2 or Google's Veo 3, and shows the evidence behind each answer: the 15 frames it sampled, the frame as the model sees it, and a probability for each of the three classes. It is available as a web app and a REST API.

Authini was founded in October 2025 by Mansur Ozaman, a computer science student at Nazarbayev University (Kazakhstan) who works on media forensics. The model is a frozen EfficientNet-B0 feature extractor followed by a support-vector machine. Every frame is degraded the way social platforms re-encode uploads (360 px, blur, JPEG quality 55, 224 × 224) before it is classified, so a clean upload and a re-encoded TikTok are judged the same way.

What Authini does not do: it only knows Sora 2 and Veo 3, so video from other generators can be classed as Real or as the nearer of the two. Its confidence is the classifier's probability, not a measured accuracy. Results under 60% are shown as inconclusive.

## Product

- [Check a video](https://authini.co/detect): upload MP4, MOV, WEBM, JPG, PNG or WEBP up to 50 MB, or paste a public TikTok link
- [Example reports](https://authini.co/examples/veo3-beach): real runs of the model on Sora 2, Veo 3 and real clips, no sign-in needed
- [Pricing](https://authini.co/pricing): Free (3 checks a month), Classic (100 checks, $8.29/month), Premium (500 checks, $17.49/month), Enterprise by agreement. Every plan uses the same model and includes the API.

## Developers

- [API docs](https://authini.co/docs): bearer-token auth, `POST /detect/` with a file or TikTok URL, JSON response with result, confidence, per-class probabilities and frames analyzed

## Research

- [Model card](https://authini.co/research): architecture, preprocessing, classes, threshold and known limitations
- [Evaluating Deep Learning and Traditional Approaches Used in Source Camera Identification](https://arxiv.org/abs/2511.19180): the founder's related paper, accepted at SIDe'26 (Springer)

## Optional

- [Full text of the main pages](https://authini.co/llms-full.txt)
- [About](https://authini.co/about)
- [Privacy policy](https://authini.co/privacy)
- [Terms of service](https://authini.co/terms)
- Contact: hello@authini.co