Audiobook Maker Pro

A desktop application for turning text, PDF, and EPUB content into audiobooks with fast, local TTS engines.

Features

• Fast local Chatterbox inference without a TTS-WebUI server
• Faster Qwen3-TTS with CUDA graphs, voice cloning, built-in speakers, and voice design
• English and multilingual Chatterbox models
• Voice-reference management
• TXT, PDF, and EPUB input
• Built-in ebook chapter editor and audio preview
• Aurora Daydream and Midnight Bloom interface themes
• Optional remote TTS-WebUI API compatibility

Interface showcase

The interface adapts its artwork and settings to the selected speech engine while keeping the audiobook workflow in one workspace.

Local Chatterbox · Midnight Bloom

Faster Qwen3-TTS · Midnight Bloom

Requirements

• Python 3.10 or 3.11
• Windows 10/11, Linux, or macOS
• An NVIDIA CUDA GPU is strongly recommended for audiobook generation

Installation

Create and activate a virtual environment first:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip

For an NVIDIA GPU, install the same PyTorch generation used by the current TTS-WebUI setup:

python -m pip install torch==2.11.0 torchvision==0.26.0 torchaudio==2.11.0 --index-url https://download.pytorch.org/whl/cu128

Then install the application:

python -m pip install -r requirements.txt
python main.py

For CPU, macOS, AMD, or Intel GPU installation commands, use the PyTorch installation selector before installing requirements.txt.

Updating an existing installation

Older versions of this repository installed a different package under the chatterbox-tts name. Remove both possible distributions once, then reinstall so that shared chatterbox module files cannot overlap:

python -m pip uninstall -y chatterbox-tts tts-webui.chatterbox-tts
python -m pip install -r requirements.txt

Usage

Local mode (recommended)

1. Select Local Chatterbox (Direct).
2. Choose chatterbox or multilingual, a precision, and an optional voice reference.
3. Enter text or open a supported file.
4. Generate the audio.

Local CUDA generation uses the optimized Chatterbox 0.4.4 inference package and its manual CUDA-graph token backend - the same performance path used by the TTS-WebUI Chatterbox extension. TTS-WebUI itself is not imported or required. The first generation can take longer while CUDA graphs and caches are prepared; later chunks reuse them.

API mode (optional)

Use API mode only when you want to generate on another machine or keep compatibility with an existing TTS-WebUI server. Activate its OpenAI-compatible API and use the default endpoint, http://localhost:7778/v1, or enter a remote endpoint.

Faster Qwen3-TTS (optional)

Qwen is intentionally installed in a separate managed environment so its PyTorch and audio dependencies cannot replace the versions used by Chatterbox.

1. Select Faster Qwen3-TTS in the TTS Engine panel.
2. Click Install Engine and let the background installation finish.
3. Choose one of the Qwen modes:
   • Voice Clone uses a reference recording and transcript. Speaker-embedding-only mode makes the transcript optional and is useful for language switching.
   • Built-in Voice uses Qwen speakers such as Ryan, Vivian, or Serena and accepts an optional delivery instruction.
   • Voice Design creates a voice from a written description.
4. Generate audio. The selected model downloads from Hugging Face on first use and is reused from the normal Hugging Face cache afterward.

The accelerated torch backend in Faster Qwen3-TTS requires an NVIDIA CUDA GPU. Its model remains loaded in a persistent worker process, so CUDA graphs and voice prompts are reused across audiobook chunks. The optional runtime is stored under the operating system's local application-data directory, not inside the project virtual environment.

Acknowledgements

• Chatterbox by Resemble AI
• TTS-WebUI Chatterbox performance fork
• TTS-WebUI for API compatibility
• Faster Qwen3-TTS by Andres Marafioti
• Qwen3-TTS by the Qwen team