---
language: en
tags:
- audio
- emotion-detection
- therapy
- wav2vec2
- custom-trained
license: mit
---

# Improved Emotional Therapy Analyzer

This model was custom-trained on therapy audio data and achieved:
- Training Loss: 1.140 (down from 1.238)
- Accuracy: 67.8% (up from 58.5%)
- F1 Score: 67.2% (up from 56.8%)

## Usage
Upload audio files for emotional analysis and strain scoring.