transcribe-anything 1.2.6.0

Input a local file or url and this service will transcribe it using Mozilla Deepspeech

transcribe-anything

Input a local file or url and this service will transcribe it using Mozilla Deepspeech 0.9.3.

Build Status

Example

• Example (cmd):
  • transcribe_anything <YOUTUBE_URL> > out_subtitles.txt
  • transcribe_anything <LOCAL.MP4/WAV> > out_subtitles.txt
• Example (api):

  from transcribe_anything.api import bulk_transcribe
  
  urls = ['https://www.youtube.com/watch?v=Erk4_jFDjzQ']
  def onresolve(url, sub): print(url, sub)
  def onfail(url): print(f'Failed: {url}')
  bulk_transcribe(urls, onresolve=onresolve, onfail=onfail)
  

Quick start

Optional: Create a virtual python package

• Works for Ubuntu/MacOS/Win32
• mkdir transcribe_anything
• cd transcribe_anything
• Download and install virtual env:
  • curl -X GET https://raw.githubusercontent.com/zackees/make_venv/main/make_venv.py -o make_env.py
  • python make_env.py
• Enter the environment:
  • source activate.sh

The environment is now active and the next step will only install to the local python. If the terminal is closed then to get back into the environment cd transcribe_anything and execute source activate.sh

Required: Install to current python environment

• pip install transcribe-anything
  • The command transcribe_anything will magically become available.
• transcribe_anything <YOUTUBE_URL> > out_subtitles.txt
• -or- transcribe_anything <MY_LOCAL.MP4/WAV> > out_subtitles.txt

How does it work?

This program performs fetching using yt-dlp for downloading videos from video services, and then stripping the audio track out.

static_ffmpeg is then called to transcode the audio track into a specific format that DeepSpeech requires.

Once the audio file has been prepared, pydeepspeech is called. This little utility automatically downloads the proper AI models and installs them into the proper path so that deepspeech can be called. It also partitions the input wav file into chunks, split at the parts of silence, in order to make processing go easier (DeepSpeech degrades performance significantly with longer audio clips, so they have to be kept short.)

Tech Stack

• Mozilla DeepSpeech: https://github.com/mozilla/DeepSpeech
• pydeepspeech: https://github.com/zackees/pydeepspeech
  • mic_vad_streaming: https://github.com/hadran9/DeepSpeech-examples/tree/r0.9/mic_vad_streaming
• yt-dlp: https://github.com/yt-dlp/yt-dlp
• static-ffmpeg
  • github: https://github.com/zackees/static_ffmpeg
  • pypi: https://pypi.org/project/static-ffmpeg/

Testing

• All tests are run by tox, simply go to the project directory root and run it.

Versions

• 1.2.6: Supports spaces in file names now.
• 1.2.5:
  • Improved handling of YouTube downloads by switching youtube-dl -> yt-dlp