SadBoys

Table of contents

• Overview
• Program Execution Flow
• Installation and Setup
• Project Components
• SadBoysCompiler Code Components
• Lessons Learned
• Future Improvements

Overview

This is a fun project done in Python. The goal of this project is to compile nostalgic/sad type of videos from the video and music inputs that the User chooses. The ideal output would be something like this video or this one. There are quite a lot of videos like that on YouTube and apparently they are quite popular - they get millions of views 😃 The project was very much inspired by the following YouTube video. While I am not into these kinds of videos myself, I simply found the idea quite entertaining and wanted to try coding the program myself using Python and to explore a little bit of its capabilities as well as get familiar with Visual Studio Code IDE 😃 😃 😃

Program execution flow

For the input original video path, offsets from the beginning and the end (in case the video contains start or end credits), desired final video length and the desired min/max durations of the video chunks, the program takes uniformly distributed cuts (subclips) from the video and stitches them together in one final video clip (there is also an option to shuffle the subclips):

Once all subclips are stitched together, the program uses the overlay background image and video to mix them on top of the sequence. Selected music replaces the original soundrack of the video. Additionally, at the beginning of the video we add a randomly selected word from the list of predefined words. Once the word is selected, it gets modified so that there are spaces between the letters.

Here is an example of what the input might look like and the produced output:

Installation and setup

Components needed for the project to work:

1. MoviePy -> can be installed by using the following command:

   pip install moviepy

2. ImageMagick

3. Optional Potentially you would need to configure Python to point to the proper ImageMagick .exe location. The config file for this action should be located in the following folder: {PathToPythonOnYourComputer}\Lib\site-packages\moviepy\config_defaults.py

   In that file you would need to replace the path in the following line with the path on your computer: IMAGEMAGICK_BINARY = os.getenv('IMAGEMAGICK_BINARY', 'C:\\Program Files\\ImageMagick-7.0.10-Q16-HDRI\\magick.exe')

Project components

The root folder contains the main executable script SadBoysCompiler that has all code of the project. NOTE: there is only one code file in this project really. I realize it could be reorganized and refactored into multiple files each of which would have its own class/functionality. For the purpose of simplicity I decided to not do that here. The assets folder include folders that contain files as follows:

• music -> folder contains music files that are going to be used as a soundtrack for the final compiled video
• overlayPics -> folder contains different pictures that are going te be used as an overlay background color theme of the final video
• overlayVids -> folder contains different short clips that are going to be used as an overlay background video theme for the final compiled video
• overlayWords -> this folder contains a file-array of words one of which is going to be used for the intro section of the compiled video
• vids -> folder contains video files that are going to be used for the final compiled video

The output folder contains compiled video examples of the program execution (as of right now these are just first program iteration passes 😃). If User chose to do so then the subsequent compiled videos would go into the same folder (or they could choose another one).

SadBoysCompiler code components

• Sequence class -> this is a bulding block class that provides the information about the time in the original video where this subclip begins and where it ends
• InputAgent class -> responsible for gathering all User inputs of original video file, music, overlay image/video, etc.
• SequenceManager class -> responsible for generation of the collection (either shuffled or not) of Sequence class objects based on the original length of the video plus User inputs about desired total length, length of a single subclip, etc.
• WordsGenerator class -> responsible for producing the word that is going to be used in the beginning of the final composed video clip
• VideoCreator class -> responsible for completing the final video compilation using output of the SequenceManager, User selected music, original video, overlay video and background image

Lessons learned

Working on this project allowed me to code in Python using external video processing library, explore its capabilities and have a glimpse of what it has to offer. Moreover, I worked in Visual Studio Code IDE, explored how to use different extensions for it and debug a Python program.

Future improvements

While working on this project I came up with a few ideas that I could explore further. Some of them are small, others would require some more time, and some of them are probably too ambitious 😆 So even though it is a pretty small project for now, it definitely has some potential.

1. Multiple overlay backgrounds and images could be used for the same video. The program could also select those automatically in a random matter, so that the User won't have to provide those inputs. Offsets could be also selected automatically - for that I would imagine we need to run some kind of video analysis (ML?) process to determine where the actual video content starts/ends. Subclip lengths could be also automatically picked.
2. It is possible to scrape YouTube for the same type of videos that were mentioned in the beginning of this readme file, download automatically and use those videos as inputs to the program. In a similar way, it is possible to scrape YouTube or any other platform for a "sad" type of free music and use it as a final video soundrack. The above would allow us to further automate the program, potentially to the point of eliminating User input altogether.
3. If the above is achieveable, we could setup this whole program to run once in a while and produce these videos automatically. These videos then would be uploaded onto various platforms like YouTube, Instagram, TikTok, etc., allowing us to have our own little "channel" with auto generated content. Whether we can upload the videos automatically remains a question for now (I believe Instagram, at the moment, doesn't have open API to actually upload videos).
4. If we scrape the videos and music, we could potentially use some kind of ML algorithm to then further analyze the results. Meaning, scraping the Internet based on tags would be just an initial data set, that will be further cleaned up by utilizing ML algorightm to narrow down the results and improve quality of our inputs (and, ultimately, quality of our outputs as well).