Relayproof

Inspiration

Videos and audio can be edited or misused, and it's not always easy to know what to trust. We built RelayProof to help people check suspicious media and understand the evidence instead of blindly trusting AI.

What It Does

RelayProof combines media verification and AI analysis in one place. It can flag suspicious moments in videos, check claims against published fact-checks and web sources, read available C2PA Content Credentials, and verify signed media claims.

How We Built It

We built it using Vite, server-side API routes, NVIDIA NIM, Google Fact Check Tools and SerpAPI. We also added speech analysis, report exports, and automated tests to catch errors and prevent API keys from leaking into the client bundle.

Challenges

The biggest challenge was making the results useful without treating AI as always correct. We also worked on handling API failures, unusual video files, and keeping credentials secure.

What We Learned

We learned that media verification needs more than AI alone. Provenance, reliable sources, and human judgement all matter. A missing credential doesn't automatically mean a video is fake, and AI findings still need review.