Is there a compression sweetspot?

[Chauceratemyhamster]

I ask because I compressed some PNG files into a video and tried two quality settings. One at 10kbps and the other 20kbps both x265. The 10kbps video just seemed to be smoother and cleaner than the larger "better quality" version. Is that all happening in my head or what

[poisondeathray]

Which one resembles the original input more closely ? That's what "better quality" means in a lossy compression context

If you did it properly - same settings for everything else - the one with 2x actual bitrate should more closely resemble the original . "Smoother and cleaner" is probably less like the original, and more detail loss ; even if you prefer smoother and cleaner


But if your PNG sequence had other subjective problems (e.g. noise) , perhaps pre-processing it would be an option or better approach .

[sophisticles]

I ask because I compressed some PNG files into a video and tried two quality settings. One at 10kbps and the other 20kbps both x265. The 10kbps video just seemed to be smoother and cleaner than the larger "better quality" version. Is that all happening in my head or what

If there was there would be no reason to have bit rate adjustments or quality settings, the encoder would just determine a "sweet spot" for each video and be done with it.

[Chauceratemyhamster]

Which one resembles the original input more closely ?

Neither, they are from an AI enhanced video so they both look better than the original video. The lower bitrate one just looks better and plays a lot smoother, by which I mean its 60fps motion is better, more natural. I've tried it with VLC and MPC and get similar results.

[poisondeathray]

Which one resembles the original input more closely ?

Neither, they are from an AI enhanced video so they both look better than the original video. The lower bitrate one just looks better and plays a lot smoother, by which I mean its 60fps motion is better, more natural. I've tried it with VLC and MPC and get similar results.

The original direct input, not the pre AI or 50 steps before... - meaning whatever your direct input is into the encoder. 2x more bitrate is going to resemble the AI output more closely

Possibly that's an AI temporal fluctuation problem that low bitrate is probably smoothing over. You should probably look into additional processing in the pipeline, instead of using lower bitrate to achieve desired results