Audio Analysis and Text-to-Speech Tool

This project provides a Python-based tool for audio analysis and text-to-speech conversion using state-of-the-art AI models.

Features

• Audio Analysis: Analyze audio files using Qwen Audio Chat model
• Text-to-Speech: Convert text to speech using multiple TTS models

Audio Analysis

The tool uses Qwen Audio Chat, a multimodal model that can understand and respond to questions about audio content. It supports:

• Analyzing any audio file (local or URL)
• Asking custom questions about the audio content
• Saving analysis results to text files

Text-to-Speech

The tool provides a flexible TTS system with multiple fallback options:

• Primary: Coqui XTTS-v2
• Alternatives:
  • Microsoft SpeechT5
  • Facebook MMS-TTS
  • Other transformer-based models

Installation

1. Clone this repository
2. Install the required dependencies:

pip install -r requirements.txt

Requirements

The following Python packages are required:

transformers
torch
soundfile
TTS (optional, for XTTS support)
sentencepiece (for SpeechT5)
numpy

Usage

# Audio analysis
analyze_audio("path/to/audio.wav", "What instruments are playing in this audio?")

# Text-to-speech
speak("Hello, this is synthesized speech")
speak("Custom voice example", voice="path/to/reference.wav")
speak("Using XTTS model", model_name="coqui/xtts-v2")

Output

All outputs are saved to the output directory:

• Audio analysis: output/analysis_filename.txt
• Generated speech: output/text_preview.wav