This project is a practical tool built for REM Waste as part of a technical interview challenge. It accepts a public video URL (MP4 or Loom link), extracts the audio, and uses AI models to classify the English accent of the speaker. It also provides a confidence score and a short explanation to assist with spoken English evaluations during hiring.
- 🎥 Accepts public video links (MP4, Loom)
- 🔊 Extracts audio using
ffmpeg - 🧠 Transcribes speech with OpenAI’s Whisper model
- 🌍 Classifies speaker’s English accent:
- American
- British
- Australian
- 📊 Provides a confidence score (0–100%)
- ✏️ Includes a brief summary of the classification
Screenshot:
The tool is designed to assist hiring managers in evaluating the clarity and regional accent of English-speaking candidates in video submissions. It can serve as an internal screening aid.
git clone https://github.com/harrisonokoth/accent-detector.git
cd accent-detector
How to run:
streamlit run app.py