GPT-SubTrans Dockerized

This repository extends and dockerizes the publicly available repository gpt-subtrans. It exposes the command line interface and uses OpenAI as the AI provider. Additionally, it adds functionality to recursively translate all .srt files under a specified folder and its subfolders.

Installation
Usage
Environment Variables
Meta File Structure
Credits
License

Installation

Prerequisites

Docker installed on your system
OpenAI API key

Build Docker Image

Build the Docker image builder stage:

docker build -t gpt-subtrans-dockerized .

Build the Docker image runtime stage:

docker build --target builder -t gpt-subtrans-dockerized .

Usage

Single File Translation

Run the following command to translate a single subtitle file:

docker run --rm -v [path to subtitles folder]:/app/subtitles gpt-subtrans /app/subtitles/[subtitle file] [args]

Recursive Folder Translation

Run the following command to translate all .srt files in a folder and its subfolders:

docker run --rm -v [path to subtitles folder]:/app/subtitles gpt-subtrans /app/subtitles [args]

Multi-threading Support

The folder-subtrans.py script supports multi-threading. Use the --max_workers optional argument to specify the number of concurrent files to process:

docker run --rm -v [path to subtitles folder]:/app/subtitles gpt-subtrans /app/subtitles --max_workers [number of threads] [args]

Essential Arguments

--moviename "movie name"
--description "movie description"
-l [language code]

All arguments and environment variables from the original repository can be used.

Environment Variables

Create a .env file with the following variables:

PROVIDER=OpenAI
OPENAI_MODEL=gpt-4o-mini
OPENAI_API_KEY=<your_openai_api_key>
MAX_BATCH_SIZE=120
# additional arguments from the original repo

Meta File Structure

When translating multiple .srt files using the folder translate option, an optional meta.json file can be added in each folder containing .srt files or only at the top folder. The folder-subtrans.py script will search for metadata in the meta.json file starting from the current directory and traversing up to the root directory. If the meta.json file is not found in the current directory, it will search in the parent directory until it reaches the root directory. If the meta.json file is not found in any of the directories, it will use the default metadata values, which are the filename as the moviename without the extension and no description.

The file should have the following structure:

{
    "file1.srt": {
        "filename": "file1.srt",
        "subtrans_args": {
            "moviename": "name"
        }
    },
    "file2.srt": {
        "filename": "file2.srt",
        "subtrans_args": {
            "moviename": "name",
            "description": "short info"
        }
    }
}

The script will translate each .srt file using the arguments passed in the RUN call and the ones in the meta.json for each .srt file. If the meta.json file does not exist when using the folder translate option, the filename without its extension will be used as the moviename argument.

Credits

This repository extends the functionality of the gpt-subtrans project. Special thanks to the original authors for their work.

License

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

OferElfassi/gpt-subtrans-dockerized