subtitle-helper

This repository contains a variety of attempts in order to find a better way to make subtitles.

After all, the most efficient way is still manual, but feel free to use the code here for more experimentation.

Please take an eye on our translation group's work

You can also find all our subtitle files here

src/

1. download.ipynb: Downloads video from YouTube link
2. extract_timestamp.ipynb: Extracts timestamps for each word
3. train_SaT.ipynb: Trains SaT using your own dataset
4. train_lora.py: When training SaT use this to replace original train_lora.py
5. whisper-timestamp.ipynb: Set up whisper-timestamped
6. whisper-timestamp.py: GUI version for whisper-timestamp.ipynb
7. whisperx.ipynb: Setup for WhisperX, only runnable on Kaggle

json/

1. lora_dummy_config.json: When training SaT, put this in target file
2. output.json: Example output of extract_timestamp
3. whisper-timestamp.json: Example output of whisper-timestamp
4. whisperx.json: Example output of whisperX

srt/

• Contains all .srt subtitle examples for a specific video

Reference

• whisper-timestamped
• whisperX
• SaT