Qwen3-TTS Studio

A professional, turnkey AI voice generator powered by Alibaba's Qwen3-TTS

Features

• Zero Configuration — One-click setup installs everything automatically (Python, dependencies, models)
• 8 Built-in Voices — English, Chinese, Japanese, and Korean speakers
• Style Control — Presets for Natural, Cheerful, Calm, Professional, and more
• Custom Instructions — Fine-tune voice emotion and delivery with text prompts
• Recording Library — Browse, play, and manage your generated audio files
• Modern Dark UI — Clean, professional interface inspired by Spotify and Discord
• GPU Accelerated — Automatic NVIDIA CUDA detection for faster generation
• Offline Ready — Works completely offline after initial setup

Requirements

• OS: Windows 10/11 (64-bit)
• RAM: 8GB minimum, 16GB recommended
• Storage: 15GB free space for models
• GPU: NVIDIA GPU with 6GB+ VRAM recommended (CPU mode available but slower)

Note: No Python installation required! The app downloads and manages its own embedded Python environment.

Installation

Option 1: Download Release (Recommended)

1. Download the latest release from the Releases page
2. Extract to any folder
3. Run qwen3_tts_studio.py

Option 2: Clone Repository

git clone https://github.com/yourusername/qwen3-tts-studio.git
cd qwen3-tts-studio
python qwen3_tts_studio.py

First-Time Setup

On first launch, the Setup Wizard will guide you through installation:

1. Python Environment — Downloads embedded Python 3.12 (~25MB)
2. AI Components — Installs PyTorch, Transformers, etc. (~3GB)
3. Voice Tokenizer — Required for all voices (~500MB)
4. Voice Model — Standard quality model (~7GB)

Simply click "Install Everything" and wait 10-20 minutes depending on your internet speed.

Usage

Basic Text-to-Speech

1. Type or paste text in the input box
2. Select a voice from the dropdown
3. Choose a style preset (or enter custom instructions)
4. Click "Generate Speech"
5. Audio plays automatically and saves to your library

Voices

Style Presets

• Natural — Clear, everyday speech
• Cheerful — Enthusiastic and happy
• Calm — Relaxed and soothing
• Professional — Confident business tone
• Excited — High energy delivery
• Gentle — Soft and tender
• News Anchor — Broadcast style
• Storytelling — Narrative delivery

Custom Style Instructions

For fine-grained control, enter custom instructions like:

• "Speak slowly with a mysterious tone"
• "Sound excited but slightly out of breath"
• "Whisper softly as if telling a secret"

File Locations

Troubleshooting

"Network error. Check your internet connection."

• Verify your internet connection is working
• Try disabling VPN if active
• Check if firewall is blocking Python
• The installer will retry automatically 3 times

Generation is slow

• With GPU: First generation loads the model (~1-2 min), subsequent ones are faster
• Without GPU: CPU mode is significantly slower, consider using the 0.6B lite model

Out of memory errors

• Close other applications to free RAM
• Use the lighter CustomVoice-0.6B model (Settings → Run Setup Wizard)
• Reduce text length per generation

Setup Wizard won't close

• Ensure all 4 checkmarks are green
• Click "Launch Studio" or "Continue to Studio"
• If stuck, close and reopen the application

Configuration

Settings are stored in studio_config.json:

{
  "selected_voice": "Ryan (English Male)",
  "selected_model": "CustomVoice-1.7B",
  "style_preset": "Natural",
  "volume": 0.8,
  "auto_play": true,
  "setup_complete": true
}

Credits

• Qwen3-TTS — Alibaba Qwen Team
• Models — Hugging Face
• UI Framework — CustomTkinter

License

This project is licensed under the MIT License - see the LICENSE file for details.

The underlying Qwen3-TTS models are licensed under Apache 2.0 by Alibaba Cloud.

Made with ❤️ for the AI voice community