X265 slow encoding

[blublub]

These are my current settings for 1080p. I don't encode yet 2160p... these are ffmpeg settings but are easily translatable to the x265 ones


ref=4:hme=0:me=star:merange=26:subme=5:bframes=8:rd=4:rd-refine=0:qcomp=0.7:fades=1:strong-intra-smoothing=1:ctu=32:qg-size=32:sao=0:selective-sao=0:cu-lossless=0:cutree=1:tu-inter-depth=4:tu-intra-depth=4:max-merge=4:rskip=1:rc-lookahead=100:aq-mode=1:aq-strength=1.0:rdoq-level=1:psy-rd=3.2:psy-rdoq=15.0:limit-modes=0:limit-refs=0:limit-tu=0:deblock=-1,-1:weightb=1:weightp=1:rect=1:amp=0:wpp=1:pmode=0:pme=0:b-intra=1:b-adapt=2:b-pyramid=1:tskip-fast=0:fast-intra=0:early-skip=0:splitrd-skip=0:min-keyint=24:keyint=240:transfer=bt709:colorprim=bt709:colormatrix=bt709

k, thx.

I use profile "default" with the following changes for UHD and a CRF of 17 via cmd:
--profile main10 --output-depth 10 --no-sao --ctu=32 --pools=64 --frame-threads=3 --qcomp=0.83 --bframes=5 --ref=5 --psy-rd=3 --hdr-opt --colormatrix=bt2020nc --chromaloc=2 --transfer=smpte2084 --no-strong-intra-smoothing --aq-mode=2 --deblock=-1:-1 --level-idc=51 --high-tier

[microchip8]

CRF 17 + qcomp 0.83 will surely bump up the bitrate a lot. But to each his own

[benwaggoner]

CRF is not a quality metric but a rate factor. It tries to maintain roughly the same factor for the whole video. It is, though, loosly tied to quality.
Specifically, RF is QP with an offset for image complexity. So it'll use lower QP for smooth gradients and higher for areas of random texture.

I've meant to do some analysis to see how well final rate factor correlates with PSNR/MS-SSIM/VMAF and subjective measurements. I wouldn't be surprised to see RF as better than PSNR at a minimum, particularly with HDR.

At least with the same settings. The biggest drawback to RF as a quality metric is it changes quite a bit with the settings, both general --preset level and more specific settings.

[blublub]

I tested a bit and between medium and slow preset the higher Bitrate at the same CRF in slow is mostly due to different settings of rd and rdoq in slow - meaning more Bitrate to keep details in motion areas instead of blurring.

BTW:
What is a better tradeoff for encoding speed and quality, increasing me or subme?

[microchip8]

I tested a bit and between medium and slow preset the higher Bitrate at the same CRF in slow is mostly due to different settings of rd and rdoq in slow - meaning more Bitrate to keep details in motion areas instead of blurring.

BTW:
What is a better tradeoff for encoding speed and quality, increasing me or subme?

--subme has less performance penalty than --me but also less quality increase. --me selects the motion estimation type (hex,umh,star,etc) while --subme is a refinement on subpixel motion

[blublub]

OK, thx. By increasing me from hex, what makes more sense umh or star?

[microchip8]

OK, thx. By increasing me from hex, what makes more sense umh or star?

umh is a motion estimation used by H.264 and implemented by x264, among others. It is a direct port to x265 and provides very decent quality. Star search is an ME "invented" for the HEVC standard and implemented by x265 among others. Star is slightly better than umh but also a bit slower. Whether you will see the difference, that is another question

I personally use star for 1 level motion estimation and umh,umh,star for HME 3 level ME.