The Problem
Understanding the “Sentiment” of video content usually requires manual watching. My client needed an automated way to gauge the emotional tone of YouTube videos and raw audio files.
The Solution
I developed Verbal, a processing pipeline that extracts audio tracks and runs them through sentiment analysis models.
- Architecture: Built with Django and deployed on DigitalOcean Droplets using Nginx as a reverse proxy.
- Core Feature: Users upload a file or paste a YouTube link, and the system generates a report breaking down emotions (Happy, Sad, Neutral, Angry).
Key Metrics
- International Delivery: Successfully delivered to a Swedish client (BothOfUs AB) as part of a remote internship.
- Full Stack: Handled everything from the UI (HTML/CSS) to the server configuration (Nginx/Gunicorn).