TTS-Narrated-Ebook-Creator

KO | EN

gitlite — search

#go #android #rust #deep-learning #linux #macos #awesome #cli #kubernetes #vue #pytorch #security #mcp #ios

TTS-Narrated-Ebook-Creator

(★ 11)

Talking books

TTS-Narrated-Ebook-Creator 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Calliope: EPUB 3 Media Overlays Creator using TTS

**Turn a text e-book into a narrated e-book with text highlighting in the EPUB 3 Media Overlays file format.**

## Key Features
* **🔒 Offline:** The method runs entirely offline, and thereby eliminating the recurring costs, privacy concerns, and copyright compliance issues associated with cloud-based solutions.
* **🎯 Zero Drift:** Calculating timestamps during TTS rather than forced alignment. The audio and text never go out of sync.
* **🎨 Layout Preservation:** Publisher’s original typography, styling, and embedded media are preserved.
* **🌓 Dark Mode Support:** Automatically generates high-contrast highlighting, e.g. purple for dark mode, yellow for light mode.
* 🧠 Supports the TTS methods **Chatterbox** and **XTTS-v2**.

---

## 🎥 See Calliope in Action

Watch a short demonstration of the system in use. This video shows the conversion process and the final playback in the Thorium reader, demonstrating the exact synchronization between the text highlighting and the TTS audio.

[**▶️ Watch the Video Demonstration**](https://youtu.be/j15BHY1hh7w)

---

## Prerequisites
1.  **Python 3.11** (Recommended) or 3.10. *(Note: Python 3.12 is not yet supported by some audio libraries).*
2.  **FFmpeg** installed on your system (required for audio processing).
    * **Ubuntu/Debian:** `sudo apt install ffmpeg`
    * **MacOS:** `brew install ffmpeg`
    * **Windows:** `winget install ffmpeg`

---

## Installation
We recommend using a virtual environment to manage dependencies. If you want to use both the XTTS and Chatterbox versions, we recommend to use different virtual environments.

```bash
# Chatterbox
# 1. Create Environment
python3.11 -m venv env_chat
source env_chat/bin/activate
# 2. Install Dependencies
pip install -r requirements/requirements_chatterbox.txt

# XTTS
# 1. Create Environment
python3.11 -m venv env_xtts
source env_xtts/bin/activate
# 2. Install Dependencies
pip install -r requirements/requirements_xtts.txt
```
---

## Usage
You need a clean reference voice sample (WAV file, ~15 seconds) to clone the narrator's voice. A sample in English is provided in assets/neutral_narrator.wav (Google translate narrator). Alternatively, several samples, for multiple languages, can be found at [Chatterbox TTS Demo Samples](https://resemble-ai.github.io/chatterbox_demopage/).

### Running with Chatterbox
Ensure your env_chat environment is active.
```bash
python src/epub_processor_chatterbox.py "my_book.epub" --voice assets/neutral_narrator.wav
```

### Running with XTTS
Ensure your env_xtts environment is active.
```bash
python src/epub_processor_xtts.py "my_book.epub" --voice assets/neutral_narrator.wav
```

### Command Line Arguments

| Argument        | Description                                                       |
|-----------------|-------------------------------------------------------------------|
| input_file      | Path to the source EPUB file.                                     |
| --voice / -v    | Required. Path to a 10-15s WAV file for voice cloning.            |
| --output / -o   | (Optional) Custom path for the resulting EPUB.                    |
| --gpu           | Force GPU usage if available (Recommended for speed).             |
| --skip-audio    | Debug mode. Generates XHTML/SMIL structure but skips TTS.         |
| --language      | Specifies the target language for the TTS engine (default: 'en')  |

---

## How It Works
Extraction: The tool unzips the EPUB and parses the internal structure (OPF/NCX).
Injection: It iterates through every HTML chapter, identifying text paragraphs. It wraps sentences in <span> tags with unique IDs without disturbing surrounding HTML/CSS.
Synthesis: It sends the text to the local AI model. The model generates audio and reports the exact duration.
Alignment: The tool writes a SMIL file linking the HTML IDs to the audio timestamps. Gaps between sentences are mathematically stitched to ensure continuous highlighting.
Packaging: The modified files are re-zipped into a valid EPUB 3 file.

---

## Troubleshooting
"FFmpeg not found" Ensure FFmpeg is installed and added to your system PATH. Run ffmpeg -version in your terminal to verify.<br>
"Out of Memory" (CUDA) Neural TTS is heavy. If you run out of VRAM, the script will automatically fallback to CPU (slower, but works).

---

## License & Acknowledgements
**Code:** The source code in this repository is released under the **MIT License**. See `LICENSE` for details.

**Models:**
* **Chatterbox:** Uses the MIT license.
* **XTTS-v2:** Released under the **Coqui Public Model License (CPML)**.

## Citation

If you use this code for your research, please cite the following paper:

> Hammer, H. L., Thambawita, V., & Halvorsen, P. (2026, April). Calliope: A TTS-based Narrated E-book Creator Ensuring Exact Synchronization, Privacy, and Layout Fidelity. In *Proceedings of the ACM Multimedia Systems Conference 2026* (pp. 416-422).

**BibTeX:**
```bibtex
@inproceedings{hammer2026calliope,
  title={Calliope: A TTS-based Narrated E-book Creator Ensuring Exact Synchronization, Privacy, and Layout Fidelity},
  author={Hammer, Hugo L and Thambawita, Vajira and Halvorsen, P{\aa}l},
  booktitle={Proceedings of the ACM Multimedia Systems Conference 2026},
  pages={416--422},
  year={2026}
}
```

TTS-Narrated-Ebook-Creator

TTS-Narrated-Ebook-Creator 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

TTS-Narrated-Ebook-Creator

TTS-Narrated-Ebook-Creator 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README