A thesis about trusting what you watch

OriginAI was built by four Computer Science students at the University of the East, Manila, so that anyone, not just experts, can check whether a video of a person has been faked.

Why we built it

Making a convincing fake video no longer takes special skills. Checking one still does. Most detection research stays in papers and code that only specialists can run.

We wanted something simpler: upload a clip, get an answer you can understand, and know how much to trust it.

A frame from a video of a man talking in a studio, with a box drawn around his face.
A frame from the video
The same man's face, cut out and enlarged.
The face it finds
The same face with a pink glow over the eyes, mouth and cheeks.
The spots it checks

From a Celeb-DF v2 clip. The glow is a simple illustration of where a detector looks.

What we promised ourselves

Four rules shaped every screen of OriginAI.

Plain words

You shouldn't need to know anything about AI to understand your result.

Show the context

Every score comes with the detector used and the line it was measured against.

Care for your video

Uploads are deleted as soon as the check ends. Results disappear after an hour.

People decide

OriginAI points you to what to look at. What's true is still your call.

Meet the team

Four Bachelor of Science in Computer Science students at the University of the East, Manila.

  • Eugene T. Custodio

    Researcher

  • Mary Kristine Degamo

    Researcher

  • Keith Roger C. Escueta

    Researcher

  • Dei Andrea E. San Diego

    Researcher

Have a question for us?

We'd love to hear from you, whether it's about a result, the detectors or the research behind them.