TruthSense

A multimodal AI system for evaluating communication skills using audio, video, and LLM-based feedback.

---

Features

• Audio analysis (transcription, speaking rate, pauses, fluency metrics)
• Video analysis (eye contact, posture, gesture, alignment)
• Fusion layer combining multimodal features
• LLM-powered structured feedback (via Groq API)
• Validated output schema for frontend integration

Dependencies

• Python 3.9–3.10
• Groq API SDK (requires GROQ_API_KEY)
• Core libraries: librosa, parselmouth, pydub, soundfile, nltk, numpy, pydantic, joblib, groq, mediapipe, opencv-python

---

Running

1. Clone the repository

git clone https://github.com/<GadiMahi/truthsense-ml.git
cd truthsense-ml

2. Create a virtual environment

python3 -m venv .venv
source .venv/bin/activate   # macOS/Linux
.venv\Scripts\activate      # Windows

3. Install dependencies

pip install -r requirements.txt

Contributors

Mahi Gadi

Utkarsh Malaiya

Made with ❤ by GDSC-VIT