all i want is a video compression algorithm that detects which parts of the image have some kind of high-frequency shit in them, stores that before it encodes the image, and then on the other side it resynthesises a kinda plausible distribution of noise/grain/dithering at the approximately correct magnitude in the roughly correct places

all i want is that. and for youtube and everyone to start using it

(bonus points if you can also resynthesise water ripples)

@jk this is basically the video version of how the opus audio codec handles "noise" (which honestly is pretty cool)

AV1 has film grain simulation: norkin.org/pdf/DCC_2018_AV1_fi which is part of the way there. it might help, at least once the encoders get good.

@jk audio coding be like:

A track with pure tones & harmonics:
FLAC: ah, nice patterns. I can make this tiny
OPUS: gah, there's nothing here i can remove, people will notice any change.

A track with lots of noise:
FLAC: uh, I guess I'll just store the uncompressed PCM
OPUS: ok, just stick some noise (doesn't matter what) in frequency band y at power x. nobody will be able to tell the difference.

Follow

@kepstin MP3: i just gotta do something. but im physiologically incapable of making noise so uhhhh. i can make some noisES. but they WILL be cosines. im sorry. lets hope you set the bitrate high enough for me to make lots of them

Sign in to participate in the conversation
Mastodon

Server run by the main developers of the project 🐘 It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!