Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/maxtheman/audiobook
Creating audiobooks from pdfs
https://github.com/maxtheman/audiobook
Last synced: about 2 months ago
JSON representation
Creating audiobooks from pdfs
- Host: GitHub
- URL: https://github.com/maxtheman/audiobook
- Owner: maxtheman
- Created: 2023-12-21T01:59:13.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2024-08-13T00:58:01.000Z (5 months ago)
- Last Synced: 2024-08-13T07:29:44.306Z (5 months ago)
- Language: Jupyter Notebook
- Size: 24.5 MB
- Stars: 0
- Watchers: 2
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Audiobook Conversion Tool
This tool converts a PDF document into an audiobook. It uses a combination of OCR, image interpretation, and text-to-speech (TTS) technologies to generate an audio file for each page of the document. The tool also provides a server for downloading the generated audio files.
## Installation
Clone this repo and create an environment from requirements.txt
The first time you use modal it will prompt you to log in. You will need to set the OPENAI_API_KEY in modal secrets for this to work.
Running audiobook.py tts requires a model to be downloaded that is ~2GB. It will be saved in the repo itself, but is ignored by git.
##Usage
Fill out chapters.yaml with the chapters you actively want converted, going off the zero-indexed page numbers.
To start the conversion process, run the following command:
```bash
modal run convert.py
```
To download files run:
```bash
modal run server.py
```
take the urls generated and replace the ones you see in `api.py` (those are mine and won't work for you)
Then run:
```bash
python api.py
```
You may to edit the functionality in `api.py` as appropriate to interact with the api.
## Directory Structure
The project uses a specific directory structure for input and output files:
- `mount/`: This directory should contain the input PDF file.
- `outputs/`: This directory is created by the scripts and contains the generated audio files and TTS models.
- `outputs/models/`: Contains the downloaded TTS models.
- `outputs/pages/`: Contains directories for each page of the PDF, each with its own audio and image files.
Please ensure that the `mount/` directory exists and contains the input PDF file before running the scripts.
## Note
The conversion process can take a significant amount of time, especially for large PDF files. The image interpretation is expensive — a full 400 page audiobook could cost $100, $80 in GPT cost from vision and $20 in GPU cost on spot pricing.
This could probably be made cheaper with some optimization, including local llama models (which this has an integration with via Ollama, but is optional.)