FactAI – Content Reliability Checker
FactAI – Content Reliability Checker
Automated analysis and fact-checking application. Accepts media files, YouTube URLs and other sources, extracts and transcribes content, structures information, then uses LLM APIs to verify the reliability of claims.
Project Objective
FactAI is a fact-checking application that works from media files (audio/video) or YouTube URLs. The user provides a YouTube link or uploads a local file, the application transcribes the content, extracts key information, then displays a readable verification table with a reliability score.
Methodology
Ingestion
The user submits a YouTube link or uploads a media file (audio/video) through the React interface. The request is sent to the Python Flask backend (POST /api/transcribe).
Processing & Transcription
Media is converted to WAV via ffmpeg. For YouTube sources, audio is downloaded via yt-dlp. Transcription is performed via Tafrigh + Wit.ai, with intelligent sentence segmentation to structure content into coherent ideas.
Verification & Score
Each extracted claim is verified via Azure OpenAI. Results are displayed in a clear table: Information, Verification, Description (with sources). An overall reliability score is computed and displayed.
Stack technique
