platput / pysubs

api to get audio transcription for video files from youtube, aws s3 and such. using OpenAI Whisper

PySubs API - OpenAI powered subtitle generator API

PySubs API is a subtitle generator API powered by OpenAI's Whisper framework.

And we are currently in beta stage. This tool currently supports YouTube video links which means if you input a video url from YouTube, it will generate subtitles and let you download the subtitle in SRT format.

I will be fixing the bugs and keep on adding more features to it until it becomes a good enough tool to go into production stage. If you have any thoughts on the tool like feedback, feature requests etc. feel free to create an issue here.

Dependencies

Firebase Dependencies

• Setup firebase project with Firestore
• Set the GOOGLE_APPLICATION_CREDENTIALS env variable with the Google service account key filepath so that PySubs can authenticate with Firebase.
• Use the CloudFunctions from pysubs-cloud-functions directory to sync the users between Firebase Users and Firestore users Collection
  • Users collection users need to have at least one credit for the users to be able to generate subtitles.
  • Currently seconds to credit ratio is based on the constant SECONDS_PER_ONE_CREDIT

Run using docker

• If running from GCP Cloudrun, add the role of Firebase Firestore Admin
• Or set the GOOGLE_APPLICATION_CREDENTIALS env variable with the Google service account key filepath so that PySubs can authenticate with Firebase

docker build -t pysubs .
docker create --name pysubs pysubs
docker start

Run using poetry

sudo apt udpate
sudo apt install ffmpeg curl
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
python3 -m poetry install
uvicorn pysubs.main:app --host="0.0.0.0" --port=8080 --log-level="info"

Testing

python3.10 -m pytest
coverage run -m pytest -v tests/ && coverage report -m

Environment

GOOGLE_APPLICATION_CREDENTIALS="pysubs.json"

Features

• Add YouTube video support
• Add firestore as the datastore to store the history of subtitle generations
• Add firebase bearer token authentication
• Add support for detecting language
• API endpoint for checking subtitle generation status
• Basic URL verification as a security measure, to be improved later
• Add api endpoint to get history
• Add credits feature
• Add firebase cloud functions to automate user sync to firestore
• Reduce credit by one for each subtitle generation
• Add limits to video duration based on available credits
• Show error message when video url is invalid and user clicks the generate button
• Add support for uploading videos
• Feature to show the details of the subtitle which is being generated
• Feature to disallow user to generate subtitles if one is being generated
• Check if the video id already exists before starting to generate the subtitle
• Add api endpoints support for subtitle translate option to different languages
• Add support for direct links of videos
• Change SECONDS_PER_ONE_CREDIT from pysubs_manager.py to use an ENV variable
• Increase test coverage