Coqui TTS Setup Guide

#Overview

This guide will help you set up Coqui TTS (Text-to-Speech) integration with your ADI dashboard. Coqui TTS provides high-quality, offline speech synthesis using advanced neural models like Tacotron2 and HiFiGAN.

#01Python Environment Setup

#1. Create Python Virtual Environment

First, create a dedicated Python environment for Coqui TTS:

cd ~
mkdir coqui-tts
cd coqui-tts

# Create virtual environment with Python 3.11
pyenv virtualenv 3.11.9 coqui-tts311
pyenv activate coqui-tts311

#2. Install Coqui TTS

Install the TTS package and its dependencies:

# Install Coqui TTS (this will take several minutes)
pip install TTS

# For CPU-only systems, install CPU-optimized PyTorch
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu

#3. Test TTS Installation

Verify the installation works:

tts --text "Hello, ADI is online and speaking through Coqui TTS." \
    --model_name tts_models/en/ljspeech/tacotron2-DDC \
    --out_path test_output.wav

# Test audio playback (optional)
aplay test_output.wav

#02System Integration

#4. Create System-Wide TTS Access

Create a symlink so the web server can access TTS:

# Find the TTS binary path
which tts
# Should show: /home/jedi/.pyenv/versions/3.11.9/envs/coqui-tts311/bin/tts

# Create system-wide symlink
sudo ln -s /home/jedi/.pyenv/versions/3.11.9/envs/coqui-tts311/bin/tts /usr/local/bin/tts

# Verify web server can access it
sudo -u www-data which tts
sudo -u www-data tts --help

#5. Set Up Directories and Permissions

Create necessary directories with proper permissions:

# Create TTS output and cache directories
sudo mkdir -p /tmp/adi_tts
sudo mkdir -p /tmp/tts_cache
sudo mkdir -p /tmp/tts_config
sudo mkdir -p /tmp/matplotlib_config

# Set ownership and permissions
sudo chown -R www-data:www-data /tmp/adi_tts
sudo chown -R www-data:www-data /tmp/tts_cache
sudo chown -R www-data:www-data /tmp/tts_config
sudo chown -R www-data:www-data /tmp/matplotlib_config

sudo chmod 755 /tmp/adi_tts
sudo chmod 755 /tmp/tts_cache
sudo chmod 755 /tmp/tts_config
sudo chmod 755 /tmp/matplotlib_config

# Create log file
sudo touch /tmp/adi_tts.log
sudo chown www-data:www-data /tmp/adi_tts.log
sudo chmod 666 /tmp/adi_tts.log

#03Web Integration

#6. Install coqui_tts.php

Create the TTS backend in your web directory:

# Create the PHP file
sudo nano /var/www/html/ai/coqui_tts.php

Copy and paste the complete coqui_tts.php code from the provided template. The key features include:

• Environment variable setup to avoid permission issues
• Text cleaning and preprocessing
• Audio generation with proper error handling
• Base64 encoding for web delivery
• Health check endpoints

#7. Update index.php JavaScript

Ensure your dashboard JavaScript calls the correct TTS endpoints:

// Update these lines in your index.php JavaScript:

// For TTS generation:
const response = await fetch('ai/coqui_tts.php', {

// For health checks:
const response = await fetch('ai/coqui_tts.php?health');

#04Testing and Verification

#8. Test the Health Endpoint

Verify the TTS system is properly configured:

curl http://192.168.1.13/ai/coqui_tts.php?health

#9. Test TTS Generation

Test actual speech synthesis:

curl -X POST http://192.168.1.13/ai/coqui_tts.php \
  -H "Content-Type: application/json" \
  -d '{"text":"Hello test","return_audio":true}'

#10. Test Dashboard Integration

Finally, test the complete integration:

1. Open your dashboard: http://192.168.1.13/ai/index.php
2. Check that the voice status shows: 🎵 Coqui Ready
3. Click "TTS Settings" button
4. Click "Test Coqui TTS" - you should hear high-quality speech
5. Try sending a regular text message - ADI should speak the response
6. Test voice input by clicking the microphone button

#Troubleshooting

#Common Issues

#❌ TTS Error on Dashboard

• Check health endpoint: http://your-server/ai/coqui_tts.php?health
• Verify TTS symlink: sudo -u www-data which tts
• Check directory permissions: ls -la /tmp/adi_tts/

#Permission Denied Errors

# Reset permissions
sudo chown -R www-data:www-data /tmp/adi_tts /tmp/tts_cache /tmp/tts_config
sudo chmod -R 755 /tmp/adi_tts /tmp/tts_cache /tmp/tts_config

#Models Not Downloading

• Ensure internet connection is available
• Check that the cache directories are writable
• Monitor /tmp/adi_tts.log for detailed error messages

#Slow Performance

• First-time model download can take 30-60 seconds
• Subsequent generations should be 2-5 seconds
• Consider using smaller/faster models for production

#Log Monitoring

Monitor system logs for issues:

# Check TTS-specific logs
tail -f /tmp/adi_tts.log

# Check Apache error logs
sudo tail -f /var/log/apache2/error.log

# Check PHP errors
sudo tail -f /var/log/php*.log

#Configuration Options

#Available Models

You can use different TTS models based on your needs:

# List all available models
tts --list_models

# Popular models:
# tts_models/en/ljspeech/tacotron2-DDC (default - good quality)
# tts_models/en/ljspeech/glow-tts (faster)
# tts_models/en/vctk/vits (multi-speaker)

#Performance Tuning

Edit /var/www/html/ai/coqui_tts.php to adjust:

• $MAX_TEXT_LENGTH - Maximum text length (default: 2000 chars)
• $MAX_AUDIO_FILES - Audio file cache limit (default: 10)
• $TTS_MODEL - Default TTS model
• $VOCODER_MODEL - Default vocoder model

#Security Considerations