VeriSon

VeriSon is a full-stack web application that analyzes a music track and estimates whether it is more likely to be AI-generated or human-created.

Try the live demo

The public demo uses Vercel for the frontend and Render for the backend. The free backend instance may take a few minutes to start after a period of inactivity.

How it works

The React/TypeScript frontend sends the uploaded audio to a FastAPI backend, where the track is processed using an MFCC + SVM inference pipeline before returning the estimated origin.

Tech stack

Frontend: React, TypeScript, Vite
Backend: Python, FastAPI, scikit-learn, librosa
Testing: pytest, Vitest, Testing Library
Tooling: Docker, GitHub Actions

Run locally

Backend

From the repository root:

docker build -f backend/Dockerfile -t verison-backend .
docker run --rm -p 8000:8000 -e CORS_ALLOWED_ORIGINS=http://localhost:5173 verison-backend

API documentation:

http://localhost:8000/docs

Frontend

cd frontend
npm ci

Create frontend/.env with:

VITE_API_BASE_URL=http://localhost:8000

Then run:

npm run dev

Open http://localhost:5173.

API

Method	Endpoint	Description
POST	/api/v1/analyze	Analyze an uploaded audio file
GET	/api/v1/model	Get model metadata
GET	/health	Service health
GET	/ready	Inference readiness

Tests and CI

# Backend
python -m pip install -r backend/requirements-dev.txt
python -m pytest backend/tests

# Frontend
cd frontend
npm ci
npm run check

GitHub Actions runs backend tests, frontend checks and a Docker build on pushes and pull requests to main.