/jp/ - Otaku Culture

>>45307535
I've done many as well myself sir, I recently trashed nearly 3 TB worth of Afestabloat, kek. The videos where FANZA's VP9 ends up bigger are far and few, most of them end up significantly smaller. A handful from Alice JAPAN, and a whole fuckton from SOD around 2021 when they had a moron either behind the camera or on the cutting room floor. It's usually more like:
>CRVR158-AbenoMiku-Takumi HEVC, Main@L5.1@High, 4096x2048, 59.940 (60000/1001) FPS, ~39.9 Mbps avg
vs.
>h_1155crvr00158 VP9, 4320x2160, 59.940 (59940/1000) FPS, ~11.4 Mbps avg
I've seen that one a dozen times. In this case, Afesta's isn't even preserving anything, x265 can and will ADD noise if you fuck with rd/psy-rd/rdoq (one of them) the "wrong" way and give it "a bunch of bits" - and it's exactly what it looks like here.
3DSVR-1088 is a bad example btw, "4KHQ" =! "Takumi" and there is actually a distinction in what Afesta means with those. IIRC:
4KHQ ~= 20 Mbps max
Takumi (ye olde) up to 40 Mbps CBR
Then they knocked it down to 30 sometime around 2020-2021. Then again to 25 shortly thereafter where it remains.
The TMPGEnc encoder they use is using x265 under the hood anyways.
FWIW, SOD's extremely noisy 1xxx-ish videos from around the same time they were angy with DMM (so Mr. Chang just ripped from AF instead) are some of the only ones where I kept Afesta's "4KHQ" versions over either FANZA AVC or VP9.
Like you said, the FANZA version technically bursts near the maximum allowable bitrate under that VP9 profile, and it's just to preserve sensor noise/some other kind of compression artifact/artificial sharpening/retard's work from somewhere along SOD's production chain.