Auto generate subtitles (.srt) given a video or audio file using WhispherX. Essentially, it takes the generated word-level timestamps and formats them according to user specifications and sentence recognization.
If you wish to download and install this:
You must first install WhispherX:
$ pip install git+https://github.com/m-bain/whisperx.git
or if WhisperX is already installed:
$ pip install git+https://github.com/m-bain/whisperx.git --upgrade
You may also need to install FFMPEG, Rust etc. Follow OpenAI's instructions here https://github.com/openai/whisper#setup.
Next, download the auto subtitles generator:
$ git clone https://github.com/lectern/AutoSubtitlesGenerator.git
$ cd AutoSubtitlesGenerator
$ python main.py
Note that in some cases, CUDA will not be installed with Torch. In which case you should do the following if you have a supported GPU:
$ pip uninstall torch
$ pip cache purge
$ pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116
You will be prompted with 3 inputs:
- file path (video|audio): relative or complete file path for any supported filetype which can be found by performing
ffmpeg -formats - no sound filter delay: the amount of no speech delay between words to consider as a pause (float > 0)
- max number of words per subtitle: the maximum number of words per each subtitle (int > 0)
Once all inputs are valid and set, a .srt file will be found in the same location as your video once the processing is done.
Feel free to contact me on discord @ lectern if any problems arise.