1 Qwen3-TTS GPU Suite

Qwen3-TTS GPU Suite

Overview

Markdown-to-Audio conversion using Alibaba’s Qwen3-TTS, optimized for NVIDIA GPUs. All model weights are stored locally, ensuring privacy and offline functionality.

GitHub Repository: https://github.com/prathmeshnik/qwen3-tts

This suite provides two main approaches:

1. Voice Cloning - Clone any voice from a reference audio sample
2. Preset Voices - Use high-quality built-in voices without reference files

Key Features

• GPU-accelerated inference using PyTorch SDPA (Scaled Dot Product Attention)
• Adaptive token budgeting for faster short-text processing
• Voice cloning capabilities with reference audio
• Multiple preset voice options (Serena, Vivian, Uncle_Fu, Ryan, Aiden, Ono_Anna, Sohee, Eric, Dylan)
• Batch processing for improved throughput
• Optimized for consumer GPUs like RTX 3050

Project Structure

See 08-directory-structure for complete file organization.

Quick Start

1. Install dependencies: pip install -r requirements.txt
2. Choose your model approach:
   • Voice Cloning: Use md_to_audio_base.py with reference audio
   • Preset Voices: Use md_to_audio_custom.py with built-in speakers
3. Generate audio from your Markdown files

Related Documentation

• 02-setup-guide - Detailed installation and hardware requirements
• 03-markdown-to-audio-pipeline - How the conversion process works
• 04-performance-optimization - GPU utilization tips for RTX 3050
• 05-cli-reference - Command-line interface options
• 06-models - Information about the Qwen3-TTS models used
• 07-integration-guide - Steps for integrating with other applications
• 09-references - External resources and model sources