About CineTruth AI

Inspiration

AI-generated and manipulated media is becoming harder to identify. CineTruth AI was built to help users analyze images, videos, and media URLs for possible manipulation using multiple forensic signals.

What it does

CineTruth AI supports:

Image analysis
Video analysis
Multiple media uploads
Media URL analysis
Visual, Audio/AV, and Context analysis
Manipulation risk score
Visual and Audio consistency percentages
Graph-based results
Professional PDF forensic reports

How we built it

The project is built with:

Python + Streamlit
Google Gemini for multimodal analysis
OpenCV for video frame extraction
Cloudflare R2 for media storage
ClickHouse for telemetry and scan records
yt-dlp + FFmpeg for supported online video URLs

Challenges

We faced challenges with video formats, cloud deployment, API limits, media URL downloading, Gemini quota/errors, and moving media from local storage to Cloudflare R2.

What we learned

We learned how to build a multimodal AI pipeline, manage cloud media storage, process videos efficiently, handle API failures, and present technical forensic results in a simple way.