>>103297650
>no 2 pass encoding

>>103297664
also that webm doesn't work btw

>>103297676
nvm it works in VLC
carry on

File: out.webm (1.29 MB, 1920x1080)

1.29 MB WEBM

This is the result of all that tranny tinkering.

MIN SSIM: 0.976 AVG SSIM: 0.988

~350 kbps

>>103297676
Works perfectly fine here
What pajeet browser are you using?

i nEEEEEEEEEEEEEEEEEEEEEED
parameters.

>>103297676
Kys poorfag

>>103297664
As of v1.15.0, Jewgle's bastard libvpx encoder CRF mode has FINALLY been fixed. Paired with -aq-mode 2 it no longer has that nasty pulsating artifacting problem anymore. Don't need to provide a bitrate ceiling anymore either too.

>>103297676
Try chrome, JUST try it. These cocksuckers also finally added support for profile 2 VP9 videos on their browser.

>>103297687
can you downscale this with libplacebo? 720p catrom_mull

>>103297691
Firefox and Chromium.

>>103297715
Just did. Didn't work neither.

>>103297754
Disable hardware accel.

>>103297718
Sorry not an MPVchad, just a lowly MPC-HC refuge who recently got into video encoding. Half the paramers I found through chatgpt and they seem to do what it says they do, especially CRF and aq-mode 2.

File: libvpx-vp9-ssim1-crf.png (76 KB, 1240x805)

76 KB PNG

>tfw achieved 4/5 MOS with all this bullshit
>literally outperforming a 2023 video codec test

AT LEAST according to SSIM thoughbeit. VMAFags, PLS comfirm 4/5 MOS. I'm doing all of this on a 2200G.

>>103297821
libplacebo (mpv shaders) is in ffmpeg https://ffmpeg.org/ffmpeg-filters.html#libplacebo
i heard its better than ffmpeg default
you can add like this

-vf libplacebo=w=-1:h=720:format=yuv420p10le:downscaler=catmull_rom

>>103297821
Both chromium flags for encoding and decoding right? Nope. It still doesn't work.

>>103297945
Don't know what to tell you my dude, I have a 2200G speed demon and Chrome displays these just fine for me. I'm also using LTSC, not sure if that matters thoughbeit. Phone/tablets render these 10-bit vp9 webms through chrome just fine too.

>>103297941
I personally wouldn't do this anyway. You're converting the native 960x540 color information into 640x360 which results in scaling issues. I hate the 720p vs 1080p wars so much, I heard some 720p anime got upscaled to 1080p by morons pretending to be competent video encoders.

>>103297965
It's a solved matter >>103297685

>>103297650
>>103297687
op looks better
what are you doing? trying to be a filesize troon or something?
bro just keep the 3.83MB file and leave it as is the 1.29MB one is trash

File: 1732178239595173.jpg (57 KB, 680x680)

57 KB JPG

>>103298062
4/5 MOS. If you want to be helpful do the VMAF shit, hotshot. We COULD be discussing 5/5 MOS via blueswordm PSY AV1 encodes but YOU KNOW.

It is what it is.

File: out2.webm (1.2 MB, 1920x1080)

1.2 MB WEBM

-tile-columns 0 -cpu-used 0

0.975 AVG SSIM: 0.987

~330 kbps

What in the god dam, this encode was slow as fucking balls and all I got was a 6% bitrate reduction?

>>103298159
Sorry for posting off topic
[spoiler]why is it that alone, among all the sexual minorities so adored by that certain cohort of people, it is only incest that has that eww gross stigma, and the dreadfully dull, boring, trite, hackneyed fucking incessant jokes about "alabama" or whatever the fuck. is it just another instance of bourgois white urban faggots hating on poor rural whites? what is the motivation and psychology behind this?[/spoiler]

>>103297650
It's funny when you realise most modern anime is produced in 720p only, and that it looks like shit regardless of how much or how little it's compressed.

>>103297676
Works for me with firefox.

>>103297650
I actually can't believe people watch this craps
Japs are fucking subhuman bug men
Might as well be collecting funko pops

File: out3.webm (1.44 MB, 1920x1080)

1.44 MB WEBM

-cpu-used 4

MIN SSIM: 0.973 AVG SSIM: 0.987

~400 kbps, roughly 15% higher bitrate. Strange, I thought it would be much higher than that. Maybe this only affects 3/5 MOS????

>>103298316
IMHO I think it's a convenient scapegoat for all the bad things happening in the world because you-know-who owns the media/hollywood. How people were conned into accepting that spreading STDs even amongst children/dogs is BETTER than some alcoholic gun-obsessed hick with a cheese brain sniffing his sister's panties is beyond me. Meanwhile you-know-who and their enemies have been cousin diddlers for centuries now lol. We're not allowed to talk about that thoughbeit.

>>103298323
Most of it isn't meant to be yosuga no sora. I mean it sucks but a lot of it is just an advertisement for something else. Even the JSDF got in on this lol, see GATE. Still there will always be a few worthy of competing with Yosuga No Sora and Oreimo (the latter being inferior OFC).

>>103298476
Beats watered down western tranny DEI shit shows any day of the week. Oreimo ending sucked mega balls but some of the core story plots were alright. I mean kuroneko in a sick twisted way did the "right" thing after all even though the male MC is a hopeless brainwashed sock puppet.

>>103298541
since you seem to still be tinkering, post your latest cli command

File: out.jpg (998 KB, 7680x4320)

998 KB JPG

>>103298610
They'are all yielding less than 5% compression efficiency gains so I'm having a really hard time finding something worthwhile posting desu senpai. I really hate to admit it but chatgpt gave me some pretty balanced parameters for 4/5 MOS. I know shitting on Google is supposed to be some kind of sport on /g/ but besides dogshit encode speeds compared to svt-av1, libvpx v1.15 seems to living up to the hype among encoding nerds.

I'll try aq-mode 1/3 next.

>>103298645
good luck buddy. Just know that whatever ffmpeg command you post will be what I use for everything

>>103298677
Probalby not a good idea. Some of these might be content specific. I'm not even sure if VP9 has synthetic grain like AV1 because if it doesn't then old movies are going to be a problem. Also the cocksuckers adding MUH GRAIN during post need to be locked up in asylums. I fucking hate that shit so much.

>>103297650
i don't have 10 bit but i do use 4:4:4
unfortunately blue clover can't play these

File: rei-clone.png (1.25 MB, 1478x2160)

1.25 MB PNG

Using my own OP params I'm going to try to encode an entire episode. Each one is originally about 500MB/~3Mbps and at least to my eyes has minimal degradation so it'll be a pretty cool real world test of VP9 outside of web bums. Not that I would use this over AV1 but this is pretty amusing given how VP9 isn't actively used in the encoding scene right now. 500 frames in, only 35,500 more to go...

CPU being used: 2200G
FPS being reported: 5 lol
ETA for 25 min 1080p episode: 2 hours

>>103297676
Filtered

>cartoons

>>103297650
I still have lots to learn but it has been really fun playing with settings trying to get a "good"* quality to size ratio.

* non-dogshit

>>103299312
Chinese ones, achtually.

File: station_psnr.png (71 KB, 1920x1080)

71 KB PNG

>>103298159
>all I got was a 6% bitrate reduction?
Welcome to the encoding life. Deal with it.

>>103297992
>You're converting the native 960x540

You don't know the native resolution yourself. Encoders use wizardry in order to find it.

>heard some 720p anime got upscaled to 1080p by morons pretending to be competent video encoders.

Upscaling native to 1080p isn't retarded since that's what your viewers players as well as your will do anyway. Controlling the quality by using a high quality/suitable scaler is essential.

Most people don't use mpv and get to watch the upscaled content with bicubic.


You can use vapoursynth-preview in order to test different scalers against each other and choose the best. I did blind testing and found that i generally prefer ewa as upscale and catmull_rom as descale for anime.