VeriWise

VeriWise is a full-stack web application designed to combat misinformation by detecting manipulated media (such as DeepFakes and AI-generated images) and verifying the factuality of textual claims. It integrates a Python and FastAPI backend for AI-powered analysis with a modern Next.js frontend for a seamless user experience.

Features

• DeepFake & AI Media Detection: Analyzes images and videos to identify manipulations and AI-generated content using specialized Transformer models. The video analysis inspects both visual frames and the audio track for a comprehensive verdict.
• Automated Fact-Checking: Decomposes textual content into individual claims, assesses their check-worthiness, generates search queries, retrieves evidence from the web (via Serper), and verifies each claim using a large language model (Gemini).
• Bias Detection: Assesses textual content for potential bias, classifying it as either "Neutral" or "Biased".
• Multimodal Input: Accepts text, images, and videos. The backend can automatically transcribe media content to text for fact-checking.
• Interactive Chat Interface: A user-friendly chat application where users can submit content and receive detailed analysis results in real-time.
• AI-Generated Summary: Provides a concise, AI-generated summary of the overall analysis results, highlighting key findings and trustworthiness.

Architecture

The application follows a client-server architecture:

• Frontend (Next.js): A responsive web interface built with React and Tailwind CSS. It provides a landing page and a dedicated /chat page for users to upload files and submit text for analysis.
• Backend (Python/FastAPI): A powerful API server that exposes a central /api/analyze endpoint. This endpoint receives multimodal input, orchestrates the analysis by running different modules concurrently, and returns a structured JSON response.
• AI Analysis Modules:
  • AI_Image.py & Manipulated.py: Handle AI-generated and manipulated image detection.
  • Deep_video.py: Performs comprehensive deepfake analysis on videos, including audio.
  • fact_verify.py & factcheck/: A sophisticated fact-checking pipeline that leverages the Gemini and Serper APIs for claim decomposition, evidence retrieval, and verification.
  • biasness.py: Detects bias in text.

Technologies Used

• Frontend: Next.js, React, Tailwind CSS, Framer Motion
• Backend: Python, FastAPI, Uvicorn
• AI & Data Processing: PyTorch, Hugging Face Transformers, OpenCV, Librosa, MoviePy, Google Generative AI (Gemini), Serper API

Installation

1. Backend Setup

To run the backend FastAPI server, follow these steps:

cd backend

python -m venv venv
# On Windows
venv\Scripts\activate
# On macOS/Linux
source venv/bin/activate

pip install -r requirements.txt

# Create a .env file in the `backend` directory and add your API keys:
# GEMINI_API_KEY=YOUR_GEMINI_KEY
# GEMINI_API_KEY_MEDIA=YOUR_GEMINI_KEY_FOR_MEDIA_PROCESSING
# SERPER_API_KEY=YOUR_SERPER_KEY
# You can add up to 5 Gemini keys (GEMINI_API_KEY_2, etc.) for parallel processing.

uvicorn main:app --reload

The backend will be running at http://localhost:8000.

2. Frontend Setup

To run the Next.js frontend, open a new terminal and follow these steps:

cd frontend

npm install

npm run dev

The frontend will be accessible at http://localhost:3000.

Usage

1. Ensure both the backend and frontend servers are running.
2. Open your browser and navigate to http://localhost:3000.
3. Click "Try Now" or navigate directly to http://localhost:3000/chat.
4. In the chat interface, you can:
   • Enter text into the input box.
   • Click the image or video icon to upload a media file.
5. Press Enter or click the Send button to submit your content for analysis.
6. The results from the various AI modules will be displayed in the chat as a structured card.

Test Data

The repository includes sample video files for testing the deepfake detection capabilities. You can find them in the backend/DeepFake/Test_data/ directory.