RipBot CQ20 lower bitrate than usual

[vertigoanduminas]

I have done most of my rips at CQ20 and the bitrate has usually been 8-10Mbps, but with my last two discs (have since updated ripbot and a few months of normal computer use/changes) I'm getting sub-5Mbps. At CQ18 the bitrate shoots past 10-12MBps which is really too high for something like Ninth Gate, etc.

Here are my logs, etc. Any ideas, or a way to use CQ19 in RipBot? Thanks:

VERTIGGO NINTHGATE ripbot profile:
--level 4.1 --vbv-bufsize 50000 --vbv-maxrate 50000 --filter -6,-5 --ref 8 --bframes 3 --b-adapt 2 --subme 9 --aq-mode 1 --trellis 2 --partitions all --me umh

[JobSettings]
VideoFile=K:\ninthgate.mkv
VideoCodec=h264
VideoWidth=1920
VideoHeight=1080
VideoAspectRatio=16:9
VideoProfile=[HIGH 4.1] VERTIGGO NINTHGATE
SourceFPS=23.976
MovieFPS=23.976
FPSinGUI=23.976
SourceNumberOfFrames=191685
SourceMovieLength=7994.86986986987
CurrentMovieLength=7994.86986986987
SourceAudioLength=0
Frames=191685
Duration=2:13:14.87
Mode=CQ
kbps=20
AudioFile=------------[NO AUDIO]------------
AudioFileSize=0
AudioCodec=unknown
AudioBPS=
AudioProfile=
Frequency=
Channels=
DurationAudio=
AudioLength=0
Normalize=OFF
AudioDelay=0
AudioLang=Undetermined    [und]
Output=J:\ninthgate.mkv
Container=mkv
LockSize=0
Size=974
Deinterlace=Do not deinterlace
Crop=Manually
CropTop=132
CropBottom=134
CropLeft=0
CropRight=0
Resize=Do not resize
ResX=1920
ResY=1080
Decimate=Do not decimate
Colors=Do not change colors
hue=0
sat=1
bright=0
cont=1
subtitles=Do not use subtitles
subtitleFileSize=0
path=
Denoise=Do not denoise 
AssumeFPS=Do not change

#MT


#VideoSource
 
video=DirectShowSource("K:\ninthgate.mkv",audio=false).ConvertToYV12()

#Deinterlace



#Decimate



#Crop
video=Crop(video,0,132,-0,-134)


#Resize



#Colors



#Denoise




#Subtitles



#AudioSource



#DownMix



#Delay



#Tempo



#Normalize



#AudioDub



#Triming



#AVSameLength



#ColorSpace

return video

I could use 2-pass, but I really prefer to use CQ for all of my encodes. It lets me sleep at night

[jagabo]

Does the picture look ok? Then why worry about the bitrate?

Some material compresses better than others. More motion, more noise, more detail, and more contrast will require more bitrate. The whole point of constant quality encoding is to ignore bitrate and just encode the video with the requested quality. Try an hour long, noiseless still frame. It will turn out with a tiny bitrate. Try a very noisy off-air analog video capture. It will require a huge bitrate.

[vertigoanduminas]

The picture looks OK, in fact, to most people it would appear identical, but I'm seeking perfection, or at least improvement of my encoding skills.

Here is the original from my .m2ts


Here is the encode at CQ18 and -6 deblock strength


Here is the encode at CQ20 and -6 deblock strength


It's close, but flipping between frames you can tell that the picture has lost some depth (i.e. dark specks are lighter and light specks are dimmer (in effect, it's as if the entire frame has been subject to the slightest blur)

I know you're right about some streams compressing better than others, but 5Mbps is just too low for 1080p picture IMO.

[jagabo]

If you go back to one of your older CQ=20 conversions that came out with a higher bitrate and compressed again with CQ=18 you will see the same improvement.

[poisondeathray]

Here are my logs, etc. Any ideas, or a way to use CQ19 in RipBot?

Type the number in the box. You're not limited to the drop down menu presets

There are also differences between x264 versions between different ripbot builds. That is the main reason for behaviour differences. The CRF (CQ) values get rebalanced and are not necessarily "equivalent" between builds. They are arbitrary, so don't get fixated on a set CRF value, because you can be certain this will change again in the future (they've been rebalanced several times now)

[vertigoanduminas]

Thanks, that was stupid of me. Anyway, using CQ16 (~16Mbps) gives exactly the same frame. I'm convinced something else in RipBot's script is affecting the image, but I don't know what it would be. I see things up there like "sat=1" and "cont=1" but they're not available to change in RipBot, which scares me. Of course they could be irrelevant to this, but it also seems possible that contrast could be the only difference between the encodes...

[poisondeathray]

In avisynth , sat=1 and cont=1 are part of tweak() parameters and default is (1.0) which means no change

Any other values would change it. So eg. Tweak(sat=1.1, cont=1.1) would be higher saturation and contrast. So by your logfile, there is nothing changed from the default in those regards. You would have to check your previous encode log to compare

Other changes have been made to x264. So depending on which version you are using, there are other functions which have been changed.

[jagabo]

There is a difference between CQ16 and CQ18. But once you go below ~CQ18 the differences get to be very subtle and hard to see.

Use an AviSynth script like this to amplify the differences between two videos:

name1="video1.avi"
name2="video2.avi"

v1 = AviSource(name1)
v2 = AviSource(name2)
sub = v1.subtract(v2)
substrong = sub.levels(112,1,144,0,255)

StackVertical(StackHorizontal(v1.subtitle(name1),v 2.subtitle(name2)),StackHorizontal(sub.subtitle("D ifference"),substrong.subtitle("Difference amplified")))