What are the implications of there being a touhou compression setting in x264?
>>109328451>elite ball knowledge isn't rea--
>>109328451What are the implications of all the tech nerds chopping off their dick?
>>109328451Touhou gameplay footage was used as one of the test datasets during the development of x264, since it was well known to have bad results with existing encoders at the time.
>>109328487at last i truly see
>>109328487makes sense considering the fuckton of colourful objects flying around at every second
>>109328451The replies are a mix of vibeslopper techbros and encoder trannies. Dire.>>109328461Zoom zoom!
>>109328487This dude knows ball
>>109328859they all own you
https://www.youtube.com/watch?v=CrPLhATu9pUWhy is bullet hell so hard on encoders?
>>109328996Which one are (You)?
>>109328451If games werent designed so autistically around millions of onscreen bullets, then we probably wouldnt need this option.
my go-to torture test for video encoding back in the day was the melancholy of haruhi suzumiya opening, has a bit of everything
>>109329037Looks like random unpredictable noise.
>>109328483>What are the implications of all the tech nerds chopping off their dick?autism is always the common denominator
>>109328487In that case creating a specific path that will be executed exclusively for Touhou doesn't make any sense.
>>109329258it'd likely come in handy for CGI-heavy workloadsI could imagine it being useful for a fluid simulationbut again only useful if the encoder happens to play touhou lmao
>>109328487anime codec
>>109329037natural enemy of compression
>>109329037Video encoders all try to optimize regions and runs of the same pixels. Bullet hells, meanwhile are a million little objects at arbitrary offsets. If you could track those objects, it'd compress extremely well but any generic algorithm for smooth film or animation fails miserably.
imagine if there was a video codec that converted the picture into a bunch of moving sprites, it would be perfect for compressing 2d video game content.
>>109328483my theory is that anime brainwhashed them with yuri anime, also atheism
>>109328451the statistic still stands
>>109329037Hint: It's the same reason (most) movies can look great when encoded at a shit bitrate like 2Mb/s @ 720p+, but most any videogame footage encoded at the same settings will look shitty.
>>109328487Smart move
That's pretty much what motion vectors do
What are the specific optimizations? Is it about many small objects moving in random directions?
>>109328487I mean, a bullet hell game is the obvious choice for a stress test for codecs.
>>109329037>millions of small objects moving in all directions with 1000 different colors at different speeds>all overlayed on top of an always moving backgroundgee I wonder why
>>109329258have you never heard of Nvidia hacking their drivers for specific game or/and benchmarks to get higher (and fake) score?
>>109334255Vampire Survivor is pretty good too. I have seen streams of it completely shit themselves. A fucking sea of high contrast objects and the codec simply decides to die.
2010-11-23 04:11:51 < Dark_Shikari> also, GCI students, feel free to hang out in #x2642010-11-23 04:12:06 < Dark_Shikari> where we discuss a wide variety of topics that are, in fact, totally x264-related2010-11-23 04:12:12 < Dark_Shikari> like touhou and starcraft2010-11-23 04:17:03 < rfw> coincidentally i have touhou music on2010-11-23 04:17:39 < Dark_Shikari> I would if I wasn't watching the gomtv live starcraft stream2010-11-23 04:17:56 < Dark_Shikari> what's with this high correlation between "working on x264" and "touhou"
>>109334255Just try to encode a video of a videogame with lots of ground grass, if you can make so every grass string, then you've got the best codec and compressing algorithm ever designed.
Would this be good for streaming games where a ton of particles are going off?
Do you pronounce it toe-hoe or two-who?
>>109335102I thought the problem with grass on YouTube was the aggressive chromatic subsampling, not some specific codec
>>109335128ou (~おう) is just a long o, so the former basically.
>>109332083yeah it would be great....i miss flash shit.
>>109335102...or maybe you just have enough bitrate for it?you could even just use whichever codec you want in lossless mode and it would work, it's not really a tough challenge unless you are targeting a specific low-ish bitrate>>109335156...what? youtube uses 4:2:0 like every video codec ever, the fact they use like 1/2 mbps for av1/vp9 at 1080p60 and 20 for 4k (which is still low for that resolution) is clearly the problem.4:2:0 is only a problem for stuff like text rendering.