🎭 Multimodal Emotion Analysis with LLaMA Inference

This Flask-based web application performs emotion analysis from text, audio, and video inputs. It leverages multiple models to detect emotional states, identify underlying causes, and generate natural, human-like inferences using the LLaMA language model via Ollama.

🚀 Features

• Text Emotion Detection: Classifies emotions from user-inputted text and extracts the likely cause.
• Audio Emotion Analysis: Converts audio to a standard format, detects vocal emotion, transcribes speech, and analyzes text emotion.
• Video Emotion Detection: Extracts visual emotions from facial cues, analyzes speech audio, and integrates transcription-based emotion classification.
• LLM-Based Inference: Summarizes emotional findings in a natural, friendly tone using LLaMA via Ollama.

🛠️ Tech Stack

• Backend: Python, Flask
• Audio/Video: moviepy, pydub, custom audio_utils.py & video_utils.py
• Emotion Models: Pre-trained models for audio, text, and visual emotion detection
• LLM Integration: Local Ollama installation running llama3.2

🧪 Requirements

• Python 3.7+
• Install dependencies:

  pip install -r requirements.txt
  

• Local installation of Ollama with the llama3.2 model
• Ensure uploads/ directory exists in your project folder

📦 Running the App

python app.py

Then open your browser and go to: http://127.0.0.1:5000

📁 Project Structure

├── app.py
├── templates/
│   └── index.html
├── static/
├── uploads/
├── audio_utils.py
├── video_utils.py
├── text_utils.py
└── README.md

📌 Notes

• Update the path to your Ollama installation (ollama.exe) in generate_llama_inference()
• Make sure media files are properly handled and cleaned up
• Easily extendable to support additional models or LLMs

🧠 Credits

Made by Abisek and Pawish
Inspired by research in AI-driven emotion understanding and contextual reasoning.