youwhisper-cli is a versatile, Python-based video/audio transcription mini-tool that uses yt-dlp and whisperx (or openai-whisper, depending on the configuration) to create a single-command transcription pipeline to transcribe audio from online video sources using yt-dlp and then passing the downloaded audio on to either whisper or whisperx for audio transcribing. Transcribed text comes out the other end, by default to .txt and .srt files.

(Full list of yt-dlp supported online platforms here)

Simply put, youwhisper-cli a handy all-in-one command line tool for creating accurate subtitles and text transcripts from online video and audio sources with a single, simple terminal command. I.e.:

youwhisper <Video-URL>

or:

youwhisper -l fi <Video-URL> 

For transcriptions where the spoken language is Finnish (fi). Program defaults to English (en). You can use the -l or --lang switch for all the available languages that openai-whisper or whisperx supports and/or modify the youwhisper.ini to suit your needs for the program's default transcription language.

youwhisper-cli runs in both Linux environments as well as in Windows using Git Bash.

• Versatile Streaming Platform Support: Utilizes yt-dlp for downloading a local audio copy from a wide range of video streaming platforms, not just YouTube.
• Advanced Transcription Capabilities: Leverages whisperx for high-quality audio transcription into text and subtitles.
• Customizable Output: Supports all the transcription output formats that whisperx or openai-whisper supports, including .srt (subtitles) and .txt (plain text).
• User-Friendly Configuration: Easy configuration adjustments via the youwhisper.ini file, accommodating different user preferences and requirements.

1. To get started with youwhisper-cli, clone the repository and set up the environment.

git clone https://github.com/FlyingFathead/youwhisper-cli.git
cd youwhisper-cli

To run youwhisper-cli, you need yt-dlp and either Whisper or WhisperX installed.

2. Install yt-dlp (via pip)

pip install yt-dlp

3. Install i.e. either whisperx or openai-whisper.

This is where things might get a bit complicated; it's highly suggested to take a look at the installation instructions on the WhisperX project page and follow them step by step.

If you're willing to take the risk of blind-installing either whisperx or openai-whisper locally, you can do so with either of the following commands:

WhisperX:

pip install git+https://github.com/m-bain/whisperx.git --upgrade

(although I highly recommend you look at the conda install instructions from the WhisperX project page)

OpenAI Whisper (via pip):

pip install -U openai-whisper

OR

OpenAI Whisper (latest cutting-edge git):

pip install git+https://github.com/openai/whisper.git 

If you have installed WhisperX (or Whisper) into a conda environment following the instructions on WhisperX's project page (which is usually the recommended method in many use scenarios, so that you don't collide with dependencies and end up with a broken Python environment), you might need to add a function to your ~/.bash_aliases file that activates the whisperx environment before the script is run, here's an example on how to do it:

# `youwhisper` /// download and auto-whisper process
function youwhisper() {
    conda activate whisperx
    python $HOME/youwhisper/youwhisper.py "$@"
    conda deactivate
}

(Note that in the function above, it's assumed that your youwhisper-cli is located at $HOME/youwhisper-cli/ (a.k.a. underneath the user's home directory, in a sub-directory named youwhisper-cli.)

Run youwhisper-cli with the URL of the video from the supported streaming platform:

./youwhisper.py <Video-URL>

This uses the default language set in the youwhisper.ini config file, which is English (en) by default.

You can specify the transcription language with i.e.:

./youwhisper.py <Video-URL> [-l <Language-Code>]

Or, if you added youwhisper to your ~/.bash_aliases, you can just use:

youwhisper <Video-URL> [-l <Language-Code>]

Adjust settings like transcription model, language, and output formats in the youwhisper.ini file to tailor the tool to your needs. Here's an example configuration of the youwhisper.ini:

[whisper]
executable = whisper
model = large-v3
language = en
output_formats = all

Configuring youwhisper.ini: A Step-by-Step Guide

1. executable:

   • This specifies the transcription tool to be used.
   • In this example, whisperx is set as the executable, meaning youwhisper-cli will use WhisperX for transcription.
   • If you have openai-whisper installed instead you can replace whisperx with whisper or the path to the openai-whisper executable.

2. model:

   • This determines the specific model of WhisperX or openai-whisper to be used for transcription.
   • large-v2 is (at the writing of this) the most current model that WhisperX supports
   • large-v3 is the default recommendation when using Whisper (Nov 2023)
   • You can select different models depending on your accuracy requirements and resource availability (e.g., base, small, medium, large, or preferably large-v2 with WhisperX. At the time of writing this, openai-whisper has large-v3 available).

3. language:

   • This sets the language for transcription.
   • The example uses en for English.
   • You can change this to any supported language code as per your requirement (e.g., fi for Finnish, es for Spanish, de for German, etc).

4. output_formats:

   • This defines the formats in which the transcribed output will be saved.
   • The default is to output to all available formats
   • You can specify different or additional formats supported by WhisperX or openai-whisper (e.g., srt, txt, json) -- note however that if all is not chosen, you can only choose one format.

By customizing these settings, you can make youwhisper fit perfectly into your workflow, whether you're a content creator, researcher, or anyone in need of quick and accurate transcriptions out of online videos.

• yt-dlp -- For audio extraction from various video streaming platforms.
• whisperx (or alternatively, openai-whisper) -- For converting audio into textual and subtitle formats.

• Contributions to enhance youwhisper are always welcome! Feel free to report issues, suggest features, or submit pull requests.
• Feel free to star the repo, share feedback, or participate in discussions if this project seems to be of any interest!

(Note: on Dec 30, 2023: changed the project+repo name to youwhisper-cli from youwhisper to avoid a repo name collision with an unrelated project)

• v1.06 - display set language
• v1.05 - bug fixes, additional checks
• v1.04 - added safety checks to see that yt-dlp and/or whisper or whisperx is installed before running, also added helpers to guide through the install process, should something be missing.
• v1.00 (Nov 17, 2023)

• (Maybe) add the option to use the Whisper API via OpenAI's API instead of a local model, for those with less local compute and access to OpenAI's API
• (Maybe) OpenAI API-based local summaries on the transcripts, possible online translation pipelines (i.e. via OpenAI API)

youwhisper-cli is made available under the MIT License. See the LICENSE file for more details. All the other required dependencies have their own corresponding licenses. See their licenses from their corresponding repositories.

Written by FlyingFathead, alongside my co-pilot, ChaosWhisperer.

• GitHub: https://github.com/FlyingFathead/
• youwhisper-cli repo @ https://github.com/FlyingFathead/youwhisper-cli/

• whisperX - https://github.com/m-bain/whisperX
• openai-whisper - https://github.com/openai/whisper
• yt-dlp - https://github.com/yt-dlp/yt-dlp