I am trying out
XMedia Recode 3.4.3.6 by taking a short 15 second/342 frame video and converting it. To begin with, I'm just replicating the encoder settings in the original video to familiarize myself with where the same settings are in Recode. I've found that I can set almost all the same settings in Recode and it has more settings than recorded in the original video.
However, using these settings takes over 20 minutes to reencode the original 342 frame video. And my PC is an Intel i7-4790K 4GHz with 32GB RAM. So, I am posting here to ask for suggestions on what settings I can dial down (and to what value) to improve rendering speed without compromising image quality.
Here are the current settings, [] bracketed ones are in Recode but not the original video.
General:
Codec=HEVC / H.265
[Frame rate=keep original]
rc=crf
crf=14.0
profile=Main 10
Level=4.1 High
Preset=Medium
Tune=Disabled
open-gop
keyint=500
[min. GOP size=0]
interlace=0
[Threads=0]
B-frame:
bframes=8
b-adapt=2 (Full)
[use b-frames as references]
weightb (weighted b-prediction?)
bframe-bias=0
Deblocking:
deblock=-2:-2
no-sao
no-sao-non-deblock
Rate Control:
vbv-bufsize=40000
vbv-maxrate=35000
[vbv-init=0.9]
crf-min=0.0
crf-max=32.0
ipratio=1.40
pbratio=1.30
qpmin=0
qpmax=32
qpstep=4
qcomp=0.60
cbqpoffs=0
crqpoffs=0
[q blur=0.5]
[blur complexity=20] (would both blur settings to 0 help? I don't think I need blurring.)
aq-mode=3 (enabled, auto-variance, bias to dark scenes)
aq-strength=1.00
[QP size=64]
cutree
no-lossless
[no-aq-motion]
[no-strict-cbr]
[no-rc-grain]
[const-vbv]
Motion estimation:
merange=44
me=3 (star)
subme=7 (HPEL 2/8, QPEL 2/8, HPEL SATD enabled)
max-merge=5 (In the last two tests, this changed to 2 even though it still says 5 on Recode's input box.)
temporal-mvp
weightp (enable weighted prediction in P slices?)
[enable weighted prediction in B slices]
[no-analyze-src-pics]
Analysis:
rd=6
ctu=32
min-cu-size=8
max-tu-size=32
limit-refs=0
rdoq-level=0
tu-intra-depth=1
tu-inter-depth=1
[limit-tu=0]
[dynamic-rd=0.00]
no-limit-modes
no-amp
no-rect
[no-b-intra]
no-cu-lossless
no-early-skip
no-fast-intra
no-tskip-fast
rd-refine
no-tskip
[no-ssim-rd]
rskip
Performance:
wpp
[pmode]
[pme]
Slice decision:
scenecut=40
[bias for scenecut detection=5]
rc-lookahead=70
lookahead-slices=6
[lookahead threads=0]
ref=5
[ctu-info=0]
no-intra-refresh
Spatial/Intra:
no-strong-intra-smoothing
no-constrained-intra
Bitstream:
[annexb]
[no-repeat-headers]
[no-aud]
[no-hrd]
[info]
no-temporal-layers
[hash=0]
can't set in Recode, different:
psy-rd=2.00 (Original: 1.2)
min-keyint=23 (Original: 1)
can't set, same:
b-pyramid
rdpenalty=0
psy-rdoq=0.00
signhide
qg-size=32
Ultimately, I would like to be able to reencode several videos from different sources to the same settings and join them together without compromising image quality. If there's a better program than Recode to do that, please let me know. I would prefer to be able to enter the settings in a UI than writing a script.