youtube-to-anchorfm VS node-fluent-ffmpeg

Interestingly enough part of this project is taken from https://github.com/Schrodinger-Hat/youtube-to-anchorfm whichh also converts YouTube videos into Anchor FM podcasts but does so in a Docker container in github actions.

I found a github repository, made by an Italian organization named Schrodinger-Hat 🐱‍💻, for converting a youtube video into an audio file and uploading it to anchor.fm. The upload task used puppeteer for that and I thought it was quite cool and simple, so I just forked it 😏. They are great, check them out!

In our live show (Schrodinger Hat) we had this necessity. So we built it for the open source community and we use this action on this repo.

Although for big files, `-fps` turns out to be slow. https://github.com/fluent-ffmpeg/node-fluent-ffmpeg/issues/4... is a much better solution - multiple `-ss` and `-i` on the same file. For the 14GB 4K video I tested, `-fps 1/462` was going to take ~4h, `-vf fps="fps=1/60"` was about 60 minutes, and the multiple `-ss` variant took ~35s.

there are lots of libraries for this. here's one in js/node: https://github.com/fluent-ffmpeg/node-fluent-ffmpeg

It need some coding skills, but if you have a base knowledge, it is pretty easy, you would need a hosting service (I use heroku because it has a free plan) and some libraries (here for node.js because it's what heroku accept). So I used the twit library to connect to the twitter api combined with a youtube search library to search for a video, youtube dl library to get the video download link and a ffmpeg library to download only the part I need and sometime resize the video.

This project wouldn't be possible without three wonderful open-source projects: the node-gtts library, the fluent-ffmpeg library and the youtube-to-anchorfm repository. I have known for years how wonderful our software development global community is, so I’m impressed, but not surprised 🥰.

For more nuanced arguments, we'll have to find a better way of setting them. The original fluent-ffmpeg project has a decent way of handling complex filters. You can pass in an array of ffmpeg filter strings or objects with keys that make it a little bit more intuitive. Would definitely want to try something along those lines.

If you are talking about creating thumbnails at any possible time in the video upon user interaction, then check this out. If this doesn't work for you, you could also just generate the thumbnail on the server by sending the server where the user paused and calling the ffmpeg CLI with child_process.