pyvttt

pyvttt is a simple Video-to-Text Transcriber written in Python.

Installation

This tool uses poetry to manage dependencies and packaging. To install all the dependencies simply run:

poetry install

pyvttt uses gpu support by default if it's available. There is no need of any additional dependencies for Nvidia gpus. If you have an AMD gpu you can install the ROCm dependencies uncommenting the torch and torchvision specific lines in the pyproject.toml file and running the following command:

poetry lock && poetry install

Don't forget to set the ROCm environment variable based on your system configuration. Example for ROCm (6.0.0) and AMD 6700XT gpu:

export HSA_OVERRIDE_GFX_VERSION=10.3.0

Usage

You can run the tool using poetry:

poetry run pyvttt --help

Or you can run the tool using python:

python -m pyvttt --help

Or you can run the tool directly from the directory or add it to your path:

pyvttt --help

usage: pyvttt [-h] [--verbose] [--debug] [--quiet | --no-quiet | -q] [--version] [--url URL [URL ...]] [--file FILE] [--output OUTPUT] [--stdout | --no-stdout | -s] [--threads THREADS] [--cpu | --no-cpu | -c] [--force-download | --no-force-download | -d] [--translate TRANSLATE]
              [--summarize SUMMARIZE] [--audio AUDIO [AUDIO ...]]

pyvttt is a simple Video-to-Text Transcriber written in Python.

options:
  -h, --help            show this help message and exit
  --verbose, -v         Increase verbosity. Use more than once to increase verbosity level (e.g. -vvv).
  --debug               Enable debug mode.
  --quiet, --no-quiet, -q
                        Do not print any output/log.
  --version             Show version and exit.
  --url URL [URL ...], -u URL [URL ...]
                        URL(s) of the video to download and transcribe.
  --file FILE, -f FILE  Path to file with urls to download and transcribe.
  --output OUTPUT, -o OUTPUT
                        Path to save the transcription.
  --stdout, --no-stdout, -s
  --threads THREADS, -t THREADS
                        Number of threads to use. Default is half of the available cores.
  --cpu, --no-cpu, -c   Force to use CPU instead of GPU.
  --force-download, --no-force-download, -d
                        Force to download the video even if it is already downloaded.
  --translate TRANSLATE, -l TRANSLATE
                        Translate transcription to the specified language. Default is english.
  --summarize SUMMARIZE, -m SUMMARIZE
                        Summarize transcription, you can define a summarization strength between 0 and 100. Suggested value: 90.
  --audio AUDIO [AUDIO ...], -a AUDIO [AUDIO ...]
                        Audio file(s) to process. Supported formats: m4a, mp3, webm, mp4, mpga, wav and mpeg.

Transcribe single video by url

# note the double quotes around the url
pyvttt --url "youtube_url" --output transcription.txt

Transcribe multiple videos by url

# note the double quotes around the urls
pyvttt --url "youtube_url" "another_youtube_url" --output transcriptions

Transcribe audio file

pyvttt --audio "path/to/audio/file" --output transcription.txt

Transcribe multiple audio files

pyvttt --audio "path/to/audio/file" "path/to/another/audio/file" --output transcriptions

Translate the transcription

You can choose to translate the audio transcription using the --translate option:

pyvttt --url "youtube_url" --output transcription.txt --translate it
# or
pyvttt --url "youtube_url" --output transcription.txt --translate italian

Supported languages

- arabic (ar)
- czech (cs)
- german (de)
- english (en)
- spanish (es)
- estonian (et)
- finnish (fi)
- french (fr)
- gujarati (gu)
- hindi (hi)
- italian (it)
- japanese (ja)
- kazakh (kk)
- korean (ko)
- lithuanian (lt)
- latvian (lv)
- burmese (my)
- nepali (ne)
- dutch (nl)
- romanian (ro)
- russian (ru)
- sinhala (si)
- turkish (tr)
- vietnamese (vi)
- chinese (zh)
- afrikaans (af)
- azerbaijani (az)
- bengali (bn)
- persian (fa)
- hebrew (he)
- croatian (hr)
- indonesian (id)
- georgian (ka)
- khmer (km)
- macedonian (mk)
- malayalam (ml)
- mongolian (mn)
- marathi (mr)
- polish (pl)
- pashto (ps)
- portuguese (pt)
- swedish (sv)
- swahili (sw)
- tamil (ta)
- telugu (te)
- thai (th)
- tagalog (tl)
- ukrainian (uk)
- urdu (ur)
- xhosa (xh)
- galician (gl)
- slovene (sl)

Summarize the transcription

You can choose to summarize the transcription using the --summarize and defining the summarization strength:

pyvttt --url "youtube_url" --output transcription.txt --summarize 90

Process multiple videos reading urls from a file

You can run pyvttt on multiple videos by using file with urls:

first_video_url_here
second_video_url_here
# this is a commented line that will be ignored
third_video_url_here

pyvttt --file urls.txt --output transcriptions

Use more threads

pyvttt [...] --threads 16

Development

Testing

To run the tests simply run:

poetry run test

Update `setup.py`

To update the setup.py file with the latest dependencies and versions run:

poetry run poetry2setup > setup.py

Acknowledgements

This project was generated using powerful tools, libraries and pretrained models such as poetry, pydantic, pytest, openai-whisper, pytube, huggingface, facebook/mbart, facebook/bart-cnn and more, I simply put the pieces together. Please check and support all the tools and libraries used in this project.

ZappaBoy/pyvttt