Summarise produced LLM-powered summaries of text and audio files. Got a three hour training video and wondering if it's worth your time? Summarise it!
There are tools like this around for web browsers, but this is purely CLI-based for automate such things via the command line.
It also supports summarising audio and text of any length (your OpenAI hard budget limits willing). It's BYO OpenAI key.
- from Youtube
- from an audio / video file locally
- from an already generated transcript file
To install Summarise, you can use pip:
pip install git+https://github.com/chorsley/audio_summariser`$ export OPENAI_API_KEY=<your key>`
`$ audio-summarise --input-youtube-url https://www.youtube.com/watch?v=3VEkzweBJPM`
# Summary of Transcript:
## Introduction
- The speaker claims to have received information that changed the course of their life six years ago.
- Allegations are made that the US government killed over 12 billion birds through the use of poisonous toxins dropped from airplanes over a period of 40 years.
- The purpose of killing the birds was to replace them with surveillance drone replicas disguised as birds.
- The speaker acknowledges that this may sound absurd but asks the audience to keep an open mind and be respectful.
...
`$ python scripts/run.py --input-audio tests/Armstrong_Small_Step.ogg --prompt "What is this audio? How is it historically significant?"`
# Summary of the Audio Transcript: "I'm going to step off the LM now. That's one small step for a man, one giant leap for mankind."
## Historical Significance of the Audio
- The audio is a famous quote from astronaut Neil Armstrong during the Apollo 11 mission to the moon.
- It marks a significant moment in human history, as it was the first time a human being set foot on the moon.
- Neil Armstrong's words symbolize the achievement of the entire human race, emphasizing the monumental nature of the moment.
...
`$ audio-summarise --input-transcript mytranscript.txt`
...Summarise can be used to summarise text files, audio files, and YouTube URLs. The summarisation process is the same for all three types of input, but the methods for processing the input are different.
$ export OPENAI_API_KEY=<your key>
$ audio-summarise --help
usage: audio-summariser [-h]
(--input-transcript INPUT_TRANSCRIPT | --input-audio-file INPUT_AUDIO_FILE | --input-youtube-url INPUT_YOUTUBE_URL)
[--debug] [--prompt PROMPT] [--model {gpt-3.5-turbo,gpt-4}]
Process a file path or YouTube URL.
options:
-h, --help show this help message and exit
--input-transcript INPUT_TRANSCRIPT
Transcript file to process.
--input-audio-file INPUT_AUDIO_FILE
Audio file to process.
--input-youtube-url INPUT_YOUTUBE_URL
YouTube URL to process.
--debug Enable debug logging.
--prompt PROMPT Custom, supplementary prompt to use for GPT model (e.g. "Make it understandable for a 5 year old.")
--model {gpt-3.5-turbo,gpt-4}
OpenAI model to use for summarisation.
If you'd like to contribute to Summarise, please fork the repository and create a pull request. We welcome contributions of all kinds, including bug fixes, new features, and documentation improvements.
Summarise is licensed under the MIT License. See the LICENSE file for more information.