Obscuring faces in videos

[Phoenix616]

Obscuring faces in videos

[n8fr8]

We used to do exactly that on video, with multiple moving resizeable boxes of obfuscation or redaction in video. However, it was quite slow to process and render, and unstable on many devices.

[n8fr8]

All of the infrastructure is in the app via the ffmpeg library to do this again, and even some of the old UI code, but it would require a few weeks/months of work to get it all modernized.

[Phoenix616]

Interesting. Thanks for letting me know, maybe I'll take a stab at it. Seeing as there seem to exist some other, proprietary apps which allow this in a pretty fast way (probably because of the speed of modern devices) I bet it could be usable now.

[n8fr8]

There are new hardware accelerated options built on the MediaCodec API but not sure which ones have the ability for the kind of dynamic pixel based filtering we need. Some examples:

https://github.com/natario1/Transcoder
https://github.com/MasayukiSuda/GPUVideo-android

We have also thought of some generative AI approaches to this, but nothing tangible there yet.