auto-highlighter-py 0.0.1

automatically generate clips from VODs!

automatically generate clips from VODs!

pip install auto-highligher

---

• you must have ffmpeg installed!
• python 3.11+
• free disk space of at least 2GB is recommended

usage

# analyzing a video and generating clips is easy!
auto-highlighter analyze -i "PATH/TO/VIDEO" 
# OR
python -m highlighter analyze -i "PATH/TO/VIDEO"

Use the --help option to see what else you can do! It is very customizable.

adjusting to get the best results

auto-highlighter will highlight moments of a given VOD based on how loud a specific point in the video is. By default, It is set to 85.0dB and if a moment goes past this value it will be highlighted.

However this is different across each video. So if needed, you can adjust the target decibel using -t <DECIBEL> option. If you don't know what target decibel to aim for, using the find-reference command will give you information about the average decibel of the video, and the greatest decibel found.

# find a target decibel
auto-highligher find-reference -i "PATH/TO/VIDEO"
# OR
python -m highlighter find-reference -i "PATH/TO/VIDEO"

TL:DR: use this command if the highlighter is creating too many, or too little clips. this will tell you the recommended target decibel to set.

:O how does it work?

The highlighter works by finding the loudest points of a given video. When a point
of a video exceeds a given target dB (default: 85.0dB), it counts that as a
clip and will compile that into a 30 seconds video.

All generated videos will automatically be outputted to a directory called ./highlights.
This directory will be created in the location where you called the command. You can also specifiy where the highlighter should output videos by using the --output, -o option.

You can also use another detection method with video! The way this method works is by taking the brightest moments of a video and creating a clip out of that too. You can also adjust the target luminance.

the tech behind it

Python 3.11+, Poetry (Package Management), FFMpeg (Video Conversion, and Generation)

Python is the programming language of choice for this project. It was very simple to use and allowed me to make this software very quickly. Poetry is used to easily publish this package to PyPI and use it in a virtual environment. FFMpeg is used on the command line to convert video to audio (for analysis) and to generate clips from highlights.

to-do

• Optimize decibel algorithm.
• Implement threading for clip generation.
• Add watch function, which can be used to create clips from ongoing streams.

ML-Based clip detection (AI)

Using Machine Learning as a means of clip detection can be done. Using my clips and it's luminance and wave data as a means of clip detection. May do this soon.

roadmap

• I am currently switching to rust as an alternative!
• GUI support is on the way.
• Documentation.