shitpost-video-generator

Generates an output video with overlayed audio and effects using a list of audio and video inputs.

How it works

Here's how it works:

1. Gets all video & audio sources to use from the input_video_sources/ and input_audio_sources/ directories. This is where you put your sources.
2. Gets a seed for the random number generator, which can be used if a user wants to re-generate the same exact video again.
3. Chooses n random videos where n is the number of videos the user wants to merge into the final results (e.g. if you want to merge 40 different videos, it chooses 40 random videos). The script will first to attempt to avoid duplicate videos. If the chosen video amount is higher than the available videos, then the script will add duplicate videos to meet that amount.
4. Applies a bunch of random effects to the video files, which include trimming the video to be a random short length, mirroring the video, speeding it up or slowing it down, reversing the video, as well as other effects.
5. Merges all the now effect-applied videos into one big video.
6. Applies effects to the audio files, which include trimming the audio to be a random short length. The script will also attempt to repeat audio clips less than 5 seconds to give a repetitive-like nature to the end result.
7. The merged video clip & merged audio clip is then merged together to produce the final result, which is put into the output/ directory.

Requirements

Python 3.8 or greater is required. This project has been tested on Python 3.8 and 3.11.

Installing dependencies:

1. Install the latest version of MoviePy: pip install -U moviepy
2. Workaround 2023-07-17: There is currently an incompatibility with Pillow and MoviePy's video.resize() method, so you'll have to downgrade Pillow for the time being: pip install Pillow==9.5.0
3. That's all! You should be able to run the main script just fine now: python3 main.py

How to use

Please take a look at the requirements above before following these steps.

1. Put a list of desired video sources into the input_video_sources/ directory, and a list of desired audio sources into the input_audio_sources directory. I recommend keeping the video file formats the same, but it shouldn't matter. The resolution of videos does not matter, either.
2. Run the script (python3 main.py), and choose a seed if you want to reproduce the same video. Otherwise, type 'any' to choose a random seed.
3. Choose the amount of videos you want the script to merge together to produce the final result. I recommend 30 for roughly minute-long videos.
4. The script will do the rest, and generate the final video in the root directory, which will be formatted as result_seed-<seed>_<number-of-videos>_<effects-enabled>.mp4.

Performance

This script uses moviepy - due to the amount of ffmpeg processes moviepy opens from this script, it has really high memory usage. I have seen this script use upwards of about ~2.5GB of memory for 40 random loaded videos. I don't know of a way around this.