How susheel-op/veritas-neural works
Veritas Neural is a local-first multimodal media-forensics application. The principal workflow is a user submitting text or an asset through the React interface, the FastAPI analysis endpoint dispatching analysis by modality, and the interface presenting a verdict and forensic breakdown. README capabilities include image and face signals, sparse video-frame analysis, audio spectral analysis, and text classification. The sampled backend excerpt confirms model loading and image/video processing details, but omits much of endpoint dispatch and frontend request code; those relationships are shown only where the README documents them.
Main entry points: main.py Ensemble Scoring · App.jsx React Interface
Components 12
User Interface
- React Interface
frontend/src/App.jsx
Analysis Service
- Analyze API
main.py
Forensic Analysis
- Image Analysis
main.py - SigLIP Scoring
main.py - Face ViT Scoring
main.py - EXIF Signals
main.py - Video Sampling
main.py - Audio Analysis
main.py - Text Analysis
main.py
Verdict
- Ensemble Scoring
main.py - Forensic Result
main.py
Other components
- User
Connections 17
- User → React Interface: submits asset
README.md - React Interface → Analyze API: sends analysis
README.md - Analyze API → Image Analysis: dispatches image
README.md - Analyze API → Video Sampling: dispatches video
README.md - Analyze API → Audio Analysis: dispatches audio
README.md - Analyze API → Text Analysis: dispatches text
README.md - Image Analysis → SigLIP Scoring: scores image
main.py - Image Analysis → Face ViT Scoring: scores face crops
main.py - Image Analysis → EXIF Signals: checks metadata
main.py - Video Sampling → SigLIP Scoring: scores sampled frames
README.md - SigLIP Scoring → Ensemble Scoring: contributes score
README.md - Face ViT Scoring → Ensemble Scoring: contributes score
README.md - EXIF Signals → Ensemble Scoring: contributes signal
README.md - Audio Analysis → Ensemble Scoring: contributes score
README.md - Text Analysis → Ensemble Scoring: contributes score
README.md - Ensemble Scoring → Forensic Result: produces verdict
README.md - Forensic Result → React Interface: returns breakdown
README.md