Rough Editor

Description

It’s a little program that I’m developing to help me make compilations a lot easier. It relies on subtitles which Youtube automatically generates.

How it works

You give the program a phrase (or regex) and it then searches for it in the subtitles and based on that it generates and runs a script called rip.sh which compiles the relevant parts into one folder (folder, not file; it doesn’t splice them). User can then make a compilation video from the generated clips. There is no need to go through every single video and search for the relevant phrase.

Installation

This application uses the following pieces of software:

• Bash,
• Python 3,
• youtube-dl,
• ffmpeg.

On e.g. Arch Linux you can download this application and install needed dependencies with the following commands:

sudo pacman -S --needed youtube-dl ffmpeg python
git clone https://github.com/kolaczyn/rough-edit.git

Linux and MacOS users obviously don’t need to worry if you they have Bash. If you’re on Windows I recommend using Git Bash. And don’t forget to add the dependencies to your PATH.

Included scripts

• download.sh — downloads both subtitles and videos of a Youtube playlist.
• convert.sh — converts the subtitles from .vtt to .srt.
• gen-script.py — the main program, generates clips in which the searched phrase occurs.
• clean.sh — cleans up the generated files e.g. if you want to rerun the program. It doesn’t delete the downloaded files.

How to use it

1. Use bash download.sh to download subs and videos of a Youtube playlist. Usage example: bash download.sh https://www.youtube.com/playlist?list=PL-p5XmQHB_JQ5YQYI7zI1uVSepG-0UPL- It puts subtitles into the ../subs folder and footage into the .../original folder.

2. If the subtitles are in .vtt format, just run: bash convert.sh This step is needed, because the program doesn’t support other formats.

3. Use gen-script.py to generate relevant clips. Here’s how you use that script:

   • The first argument you pass is the searched phrase, for example okay. You can also use regex, but keep in mind that I’m using grep to find phrases, which doesn’t support metacharacters like \w or \s are not supported.
   • The second argument is either fast or slow. The fast takes less than five seconds to run, but the footage it generates is of bad quality: sound is not synchronized, image is missing, etc. The slow takes on my machine roughly 1 hour per 15 hours of footage, but the generated footage is perfect. I recommend first running first fast, check if that’s what you want to render, then in slow mode.
   • The third and the final argument tell how much padding you want around the searched phrase. Keep in mind that if you pass in 5, it doesn’t mean there will be about 10 seconds. If you don’t give the third argument, it uses a default value of 10 seconds.

4. That’s it. Generated clips are inside ../output folder. They are sorted in chronological order. You can start editing. If you want to make another batch keep in mind that gen-script.py automatically runs clean.sh script which deletes the content of the output folder. It doesn’t delete the original footage, subtitles and any of your other data. Or at least it should behave this way.

Usage example

bash download.sh https://www.youtube.com/playlist?list=PL-p5XmQHB_JQ5YQYI7zI1uVSepG-0UPL-
bash convert.sh
python gen-script.py okay fast 5
# Now I check out the generated footage
# If I'm satisfied with the output, I use the slow mode
python  gen-script.py okay slow 5

Limitations

Like I said, the program heavily relies on Youtube’s auto-generated subtitles. So when the phrase you want to search for is not a word from a dictionary you have to find a workaround. You can find yourself moments when a Youtuber says the phrase and see how the subtitle generator interprets it. Keep in mind there can be a few ways the phrase is interpreted. Let’s say you want to search for a phrase ‘hermitcaft’. The sub generator interprets it as ‘hermit craft’ and ‘hermit crab’. So you could for example just search for the phrase ‘hermit’. It’s still better than nothing.

Disclaimer

I am not a professional developer. The script here may be suboptimal or it may delete all of your unprotected files. You don’t know what code a random person on the Internet wrote. So I recommend reviewing the source code before you run it. It’s not very long, it’s less than a 200 lines of code.

TODO list

• Add installation instructions for other Linux distros, MacOS and Windows.
• Make it so rough-editor istall itself in the PATH and you can e.g. type rough-editor Laughter fast 5 in the terminal and it then runs the program on the current folder.
• Allow to specify quality of downloaded videos.
• Give an option to first download subtitles, search for the phrase and only download relevant videos from the playlist. Should be relatively easy.
• Stop using grep and switch to program which lets you utilize all of xegex magic.
• Make conversion faster; fix the slow method and make it fast and reliable.
• Make the program run differently if the files are .mp4 and .mkv. Don’t remember why I wrote that.
• Handle situation if the subtitles don’t exist but the corresponding video does.
• Search for other edge cases and fix them.