[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
▼ Settings Mobile Home
/g/ - Technology
Return Catalog Bottom Refresh
Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.
08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]
[Advertise on 4chan]
[Return] [Catalog] [Bottom]
kiker !!m8ZMOnWSI0a
H264 = waste of encoding resou(...) No.108727230
File: hmmm.png (90 KB, 1349x632)
90 KB PNG
H264 = waste of encoding resources? No.108727230
Somebody else please test this on their PC and/or let me know if I fucked something up. If true encoding H264 video now officially takes longer to encode and wastes more storage space vs AV1.

This is from the Pedestrian Area 1080p25fps Y4M (~1GB): https://media.xiph.org/video/derf/y4m/
I compressed to ~130MB with x264 slow CRF 10 so very very very tiny hit to OG quality: https://marafile.cc/d/7e15af

Get-CimInstance -ClassName Win32_Processor | Select-Object Name, NumberOfCores, NumberOfLogicalProcessors
ffmpeg -v quiet -stats -i crf_10_h264.mkv -c:v libsvtav1 -crf 34 -y crf_34_av1.webm 2>&1 | Select-Object -Last 1 ;Write-Host ""
ffmpeg -i crf_34_av1.webm -i crf_10_h264.mkv -filter_complex libvmaf -f null - 2>&1 | Select-Object -Last 3 ;Write-Host ""

ffmpeg -v quiet -stats -i crf_10_h264.mkv -c:v libx264 -crf 24 -y crf_24_h264.mp4 2>&1 | Select-Object -Last 1 ;Write-Host ""
ffmpeg -i crf_24_h264.mp4 -i crf_10_h264.mkv -filter_complex libvmaf -f null - 2>&1 | Select-Object -Last 3 ;Write-Host ""


This is powershell and please rename the marafile.cc download to crf_10_h264.mkv
>>
Anonymous
 No.108727541
No.108727541
anon i5-4460 makes 19fps encode speed with cam 1080p vid h264 filesize 3:1 ish
>>
Anonymous
 No.108727542
No.108727542
comparing apples to oranges with regards to presets and crf values (and are forgetting essential flags for x264 like tune, faststart and yuv420p)

unless you're streaming, no reason to ever use less than veryslow (or even placebo if it's a short clip) for x264 if you're minmaxxing high quality and low bitrate

av1 takes too much cpu to decode, don't like fan noise (even if it's noctua cpu fan), and i'm not buying a new PC with the current RAM & SSD prices. however my 11-year-old nvidia gpu can decode vp9 and h265.
>>
kiker !!m8ZMOnWSI0a
 No.108727794
No.108727794
File: av1-battery-life.webm (3.84 MB, 480x640)
3.84 MB WEBM
>>108727542
>comparing apples to oranges with regards to presets and crf values (and are forgetting essential flags for x264 like tune, faststart and yuv420p)
Everything is 8-bit 4:2:0 and both encodes were validated with VMAF 95. I didn't select for any extra parameters because the defaults are used by most GUI encoders.

>av1 takes too much cpu to decode
Software decoding has definitely gotten much much better now, the Dav1d optimizations are ramping up. I recently bought an A16 5G, a poorfag $100-200 phone released in 2024, and it can decode 10-bit 1080p25fps ~1 Mbps video without lagging or overheating for about 8-10 hours non-stop.

I'm trying to figure out how to benchmark AV1, Android doesn't let you see % cpu utilization anymore so I can't estimate if it can handle 60 FPS.
>>
Anonymous
 No.108727827
No.108727827
>>108727794

yes encoders are better now but yet a dvd is 2-6Mbps ish
>>
Anonymous
 No.108727912
No.108727912
You dont need more than crf 17
>>
kiker !!m8ZMOnWSI0a
 No.108727945
No.108727945
>>108727912
CRF 22-26 if you keep the ceiling at 95 VMAF. I had to go down to CRF 10 to minimize how much grain denoising occurred otherwise it would have skewed the results in AV1's favor too much. The Pedestrian Area test file seems like a testing ground for AV1's synthetic grain feature, still no clue how that works desu.
>>
Anonymous
 No.108727978
No.108727978
>>108727230
H264 takes too much space and the results are shit unless you use a stupid high bitrate which defeats the purpose of lossy encoding.
AV1 saves the most space but rapes your CPU and is super slow, also blurry.
VP9 is unironically the best. Sharp, can handle low bitrates way better than h264 and even toasters are fine with it.
>>
kiker !!m8ZMOnWSI0a
 No.108727990
No.108727990
>>108727978
That's why I made this thread though, from my testing H264 will rapes your CPU and is super slow now.

Like they switched places lol.
>>
Anonymous
 No.108728329
No.108728329
>>108727990
Why are you talking like h.265 isnt the new standard? And unless youre a render farm, why are you using cpu render? Nv h.265 produces great results and a lower size. Gpu is simply the way to go for consumer and prosumer uses.
>>
Anonymous
 No.108728451
No.108728451
>>108727230
>i5-9500t 2.2ghz
>winblows
maybe you should get a job instead of shilling video encoders
>>
Anonymous
 No.108728500
No.108728500
File: 1766395731464167.png (80 KB, 1200x674)
80 KB PNG
>>108728329
retard
>>
Anonymous
 No.108728883
No.108728883
>>108727542
Did anyone ever tell you that you sound like a fag and your shit's all retarded?
>>
Anonymous
 No.108728935
No.108728935
>>108727230
>Retarded literal who chud thinks he knows more than the industry vets that made this encoder
>>
Anonymous
 No.108728996
No.108728996
>>108727230
all you need is mpeg2
>>
Anonymous
 No.108729457
No.108729457
>>108728500
that graph doesn’t have h.265 on it AFAICT
>>
Anonymous
 No.108729526
No.108729526
>>108727230
>6 core 2.2ghz cpu
did you travel in time from 2008?
>>
Anonymous
 No.108729631
No.108729631
>>108728329

most users do not care these things just use distro included stuff and if video is clean and maybe have titlepic and captionings then great
>>
Anonymous
 No.108730294
No.108730294
>>108728500
>retard
Thats h.264?
>>
Anonymous
 No.108730299
No.108730299
hvec > av1 > h264
I extensively local stream to my odin 2 portal and samsung phone
>>
Anonymous
 No.108730345
No.108730345
>>108727230
you have to learn that you have to tune the encoder per movie.
there is no "magical only turn the crf knob".
the moment there is grain and shit you are into a world of hurt and have to tune everything.
too bad with svt-av1 no matter what you do it will smudge out the grain
>>
Anonymous
 No.108730350
No.108730350
File: 1764766691943758.png (115 KB, 655x467)
115 KB PNG
>>108729457
>>108730294
It has intel AV1, which is considered the best AV1 hw encoder out there (ie beats hevc hw encoder). Preset 8 SVT-AV1 crushes it. Preset 7 SVT-AV1 wipes the floor with it. And so on and so on.

The point is hw video encoders don't have the best compression efficiency but speed. You'd have to be a complete moron to use them for anything other than streaming.
>>
Anonymous
 No.108730394
No.108730394
File: 1770752505708688.jpg (25 KB, 460x460)
25 KB JPG
>>108730345
There are pre-tuned svt-av1 encoders for autists like you. I kinda don't mind AV1 smoothing out the grain in video because I find it annoying in general, like ants crawling on my screen.

https://github.com/BlueSwordM/svt-av1-psyex
>>
Anonymous
 No.108730563
No.108730563
>>108727990
>from my testing H264 will rapes your CPU and is super slow now.
lol wtf? Its not like h264 is magically different today then from the past
>>
Anonymous
 No.108730618
No.108730618
>>108727230
>VMAF
AV1 development used VMAF as an optimization target, unlike H.264/x264 (at that time, VMAF didn't even exist). That makes VMAF a very poor basis for such a comparison, see Goodhart's law:
>When a measure becomes a target, it ceases to be a good measure.
>>
kiker !!m8ZMOnWSI0a
 No.108730925
No.108730925
>>108730618
I can believe that for the AOM AV1 encoder, there's actually specific -tune parameters in there for gaming VMAF and even VMAF NEG scores.

https://aomedia.org/docs/aomenc-command-line-parameters-draft.pdf

However SVT-AV1 doesn't have this. This is a separate encoder built for quick AV1 encodes like 5% less efficient than AOM AV1 encodes, so it drops many of the fancy stuff.

https://gitlab.com/AOMediaCodec/SVT-AV1/-/blob/master/Docs/Parameters.md

If you know of a better video quality metric that takes human weakness to motion (ie not SSimulacra2) into account then I'd love to hear about it though. All I know is PSNR/SSIM has been obsolete for measuring video quality since HEVC was released.
>>
Anonymous
 No.108731671
No.108731671
File: 1769942181584278.jpg (431 KB, 1133x1197)
431 KB JPG
>encoding speed test
>look inside
>it's cpu encoding
>using a 7 year old midrange cpu
>>
kiker !!m8ZMOnWSI0a
 No.108731764
No.108731764
File: no-good.webm (41 KB, 1056x720)
41 KB WEBM
>>108731671
I mean that's kinda the point man, encoding AV1 video no longer requires a threadripper class CPU meaning AV1 isn't going to be a niche thing anymore. AV1's biggest problem is being solved so well it's making H264 obsolete, at least as far as CPU video encoding is concerned.
>>
Anonymous
 No.108731843
No.108731843
>>108731764
>threadripper
It never needed that. The cheapest Arc GPU in the world can AV1 encode faster and better. No need for anything CPU related.
>>
kiker !!m8ZMOnWSI0a
 No.108731880
No.108731880
>>108731843
GPU video encoding is inferior to CPU video encoding. It's okay for video streaming your video gays but not for personal movie encodes.
>>
Anonymous
 No.108731921
No.108731921
File: 1757109139816089.png (478 KB, 625x352)
478 KB PNG
>>108731880
>there are video equivalents to audiophiles
I can't wait for the "w-well, if you zoom in 1000x on a 4k encode th-the gpu encode has 0.000001% more infetterence in the dodecahedron cortex" comparisons.
>>
kiker !!m8ZMOnWSI0a
 No.108732283
No.108732283
File: computer.webm (55 KB, 960x720)
55 KB WEBM
>>108731921
If I were a videophile I wouldn't be using VMAF at all. Instead I'd be treating every single video frame as an image and run ssimulacra2 comparisons instead, 25 times per second...

Anyway you're not going to get the highly sought after 1GB high quality 2-hour 1080p movie rips with GPU video encoding.
>>
JonSneeders !q710i/bPrg
 No.108732314
No.108732314
>>108727230
Install Gentoo.
>>
Anonymous
 No.108733673
No.108733673
File: 1756806514318552.jpg (159 KB, 933x700)
159 KB JPG
Jumping jesus, a thread about technology and not AI/twitter shit on /g/? Is hell freezing over?
>>
Anonymous
 No.108733728
No.108733728
>>108732283
It's trivial to get that with gpu encoding. It's not 2012 with nvenc gen 1 anymore. Like I said, audiophool nonsense leeching into the videosphere.
>>
Anonymous
 No.108733749
No.108733749
>>108733728
>trivial
Stop saying this shit you armchair frauds. I DO NOT KNOW WHAT IT MEANS.
>>
Anonymous
 No.108735441
No.108735441
vp9 every time
>>
Anonymous
 No.108735961
No.108735961
>>108727230
last i checked codecs relatively encode at 'same' quality like.
h264 100fps 90mb
h265 27fps 60mb
av1 8fps 40mb

so you did something wrong.
av1 is mostly trash because they keep updating something that should have had everything at release. the reason to encode something is to get a smaller filesize and delete the original.

av1 from 3 years ago isn't the av1 now, why bother with such a shitshow of a codec. with such a powerhungry encoding codec you can't expect people to reencode terabytes of video. it'll be cheaper to get a large hdd and not reencode anything to av1.
>>
Anonymous
 No.108735983
No.108735983
>>108727230
>convert all video I have to av1 crf 32 using preset 5
>it’s been over two weeks on 9800x3d and it’s only 30% done
av1 is many things but fast isn’t one of those
>>
Anonymous
 No.108736002
No.108736002
>>108727230
Yes but it's also easier to decode on older cpus
[Return] [Catalog] [Top]
Post a Reply
Return Catalog Top Refresh
[Advertise on 4chan]
Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.