[caryme]

I've been using ffmpeg's xgrid filter (similar to hstack and vstack, but allows for arbitrary grids) to produce virtual choir videos for my church choir during covid. Here's an example: https://www.youtube.com/watch?v=Oeg9w8X6hrA.

A lot of people are producing virtual choir videos right now, but I suspect few use a process similar to mine. I use Audacity to edit the audio separately, then crop the input videos using a face-aware cropping script (which uses https://github.com/ageitgey/face_recognition), then generate a video grid using ffmpeg + xgrid.

[iamstef]

I would be interested in seeing your script if you don’t mind sharing

[caryme]

Sure!

Here's the face-aware crop script (somewhat quick and dirty - I haven't moved the output dimensions to CLI arguments yet): https://gist.github.com/carylee/62dbe19579ed9fa1714a3635ff0b...

Here's an example script for stitching the output video: https://gist.github.com/carylee/a13ae1a1ecb2f81c6536476f6dfb...

And here's the resulting video (text was added later by a different editor): https://www.youtube.com/watch?v=f7Pm4nLPslY

[iamstef]

Nice, this was super timely as I was literally building the same thing. And I had just reached the part where I was annoyed that all the submitted videos had different shapes and sizes which as you know would be tedious to correct manually.

You for sure saved me some time!