The Ultimate Local AI Audiobook Producer
简体中文 | English
Most TTS tools are either too simple or too complex. Qwen3-TTS Studio is built specifically for Audiobook Production, combining high-end AI research with a user-friendly, portable interface.
- 🚀 Dual-Platform Turbo: Mac (M1 ~ M5) and Windows (Beta).
- 🎭 Pro Casting: Advanced multi-role orchestration. No more editing audio clips manually.
- 🎨 Creative Control: Design unique voices from text descriptions.
- 🛡️ 100% Privacy: Runs entirely on your local hardware. No cloud, no subscription fees.
- Google Drive (Recommended): Lite Version - No models included
- Quark Pan: Includes both Full and Lite versions
- Official Model Source: Qwen3-TTS-12Hz-1.7B-Base (Hugging Face)
- Download & Unzip: Download this repository.
- Run: Double-click
launch_studio.command. - Turbo Install: Automatic local environment setup (
./runtime_env) using our optimized Pip strategy.
- Run: Double-click
launch_studio_windows.bat. (Note: First run will verify/install environment automatically) - GPU Power: Automatically detects NVIDIA GPUs for high-speed CUDA inference.
- Role-Based Synthesis: Use a standard
Role: Dialogueformat. - Map-Reduce Engine: Efficiently processes long scripts by grouping character lines to minimize model swapping.
- Director's Tags: Use
[p=dur](e.g.,[p=1.5]) to insert precise silence segments.
- Generative TTS: Describe a voice (e.g., "Deep, raspy male voice with a slight British accent") and hear it instantly.
- One-Click Save: Like the voice? Save it to your
./voiceslibrary with one click and use it in your next book.
- Zero-Shot: Clone any voice from a 5-second sample with near-perfect fidelity.
Narrator: The forest was silent, save for the crackle of a distant fire. [p=0.8]
Geralt: I smell trouble.
Jaskier: You always smell trouble, Geralt! [p=0.5]
Special thanks to the Alibaba Qwen Team for open-sourcing the incredible Qwen3-TTS model. This project would not be possible without their contribution to the AI community.
- Model Source: Qwen3-TTS on Hugging Face
Biuboom Flow
- 📺 YouTube: @BiuBoomFlow_nothing
- 🚀 Support the project by subscribing for more AI tools!
- For Research Only: This project is intended for scholarly and research purposes only.
- User Responsibility: Any misuse of synthesized audio (including but not limited to fraud or impersonation) is the sole responsibility of the user.
- Copyright: The Qwen3-TTS model weights are owned by Alibaba Group.
Licensed under Apache 2.0. See LICENSE for details.