Fact Checker

• 

  Architecture Diagram

About the Project — Fact Checker

Inspiration The rise of misinformation—especially through short videos, edited images, viral audio clips, and AI-generated content—motivated us to build Fact Checker. In today’s world, anyone can create convincing multimedia, and traditional text-based fact-checking tools are not enough. I wanted to create a tool that empowers everyday users to quickly verify the authenticity of the content they consume.

What it does Fact Checker is a multi-format fact-checking web application that allows users to upload videos, audio files, images, or text and instantly receive:

• AI-generated fact-check analysis

• Verified answers grounded in live Google Search results

• Citations and evidence sources

• Extracted transcripts for audio/video

• A complete history of all fact-checks

Admins get a dedicated dashboard for reviewing user submissions and adding expert comments.

How we built it

• Backend: FastAPI (Python) for processing, authentication, and file handling

• Frontend: React + TypeScript + Vite

• AI Engine: Gemini 2.5 Flash with Google Search Grounding

• Speech Processing: Google Cloud Speech-to-Text for converting video/audio to text

• Media Processing: FFmpeg

• Storage: CSV-based lightweight database + local file storage

• Authentication: JWT with role-based access (User/Admin)

The project is fully modular, with separate backend and frontend services and clear routing, services, and model layers.

Challenges we ran into

• Integrating Google Cloud Speech-to-Text with video extraction pipelines

• Managing large file uploads and ensuring stable FFmpeg processing

• Designing a clean UI that supports multiple file types

• Handling API errors from Gemini when grounding or parsing complex outputs

• Designing both user and admin flows within one application

• Keeping everything lightweight while still offering reliable storage and history features

Accomplishments that we're proud of

• Successfully building a complete end-to-end multimedia fact-checking system

• Implementing real-time search-grounded fact verification

• Creating a clean user experience with clear citations and explanations

• Designing an admin dashboard for reviewing and commenting on fact checks

• Handling videos, images, texts, and audio with a unified workflow

• Completing a full-stack project within hackathon constraints

What we learned

• How to integrate multiple Google Cloud services with FastAPI

• Best practices for handling large file uploads and processing pipelines

• Designing multi-role authentication securely using JWT

• Implementing AI fact-checking grounded in real search data

• Frontend–backend communication for real-time status and results

• How challenging and important it is to design truthful AI systems

What’s next for Fact Checker

• Add support for URLs (fact-check web articles directly)

• Deploy to cloud infrastructure for public access

• Replace CSV storage with PostgreSQL / Firestore

• Add a browser extension for instant fact-checking

• Improve verification algorithms with cross-source consensus logic

• Introduce fake-media detection tools (deepfake detection, image metadata analysis, etc.)

• Build mobile apps for quick fact-checks on the go

Built With

• fastapi
• ffmpeg
• gemini-api
• google-cloud-speech-to-text
• python
• react+typescript+vite