Bitrate in new builds in CRF mode

[InTrancer]

i have test build 947 and 999 on the same source (anime 1002 frames) and result bitrate was very big
build 947: 703.60 kb/s
build 999: 1932.93 kb/s

build 947 command line
C:\Documents and Settings\SHiSH\Desktop\x264>x264.Build.947.exe --crf 14.0 --ref 16 --mixed-refs --no-fast-psk
ip --bframes 16 --b-rdo --b-pyramid --weightb --direct auto --subme 7 --trellis 1 --partitions all --8x8dct -
-me umh --merange 32 --threads auto --thread-input --progress --no-dct-decimate --output "I:\Work\higurashi\dv
d5\dvd2\s21.build947.mkv" "I:\Work\higurashi\dvd5\dvd2\VIDEO_TS\s21.avs"
avis [info]: 848x480 @ 23.98 fps (1002 frames)
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 PHADD SSE4 Cache64
x264 [info]: slice I:7 Avg QP:12.05 size: 37886 PSNR Mean Y:50.92 U:52.55 V:52.59 Avg:51.40 Global:51.35
x264 [info]: slice P:548 Avg QP:12.74 size: 5649 PSNR Mean Y:49.39 U:51.79 V:52.09 Avg:50.08 Global:50.04
x264 [info]: slice B:447 Avg QP:14.02 size: 703 PSNR Mean Y:49.33 U:51.63 V:51.73 Avg:49.98 Global:49.87
x264 [info]: consecutive B-frames: 44.8% 0.0% 2.1% 7.6% 9.0% 30.8% 2.8% 0.0% 1.8% 1.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
x264 [info]: mb I I16..4: 35.6% 44.1% 20.3%
x264 [info]: mb P I16..4: 1.4% 0.5% 0.3% P16..4: 53.0% 6.2% 11.4% 0.1% 0.0% skip:27.0%
x264 [info]: mb B I16..4: 0.0% 0.0% 0.0% B16..8: 15.4% 0.2% 0.3% direct: 0.3% skip:83.8% L0:45.1% L1:52.4% BI: 2.5%
x264 [info]: 8x8 transform intra:30.3% inter:38.0%
x264 [info]: direct mvs spatial:98.9% temporal:1.1%
x264 [info]: ref P L0 73.0% 10.3% 5.9% 1.8% 1.6% 1.5% 1.2% 0.5% 0.6% 0.6% 0.6% 0.5% 0.5% 0.5% 0.5% 0.3%
x264 [info]: ref B L0 76.2% 11.3% 5.1% 2.5% 1.0% 1.2% 0.7% 0.3% 0.3% 0.2% 0.2% 0.3% 0.2% 0.3% 0.1%
x264 [info]: ref B L1 93.2% 6.8%
x264 [info]: SSIM Mean Y:0.9930943
x264 [info]: PSNR Mean Y:49.373 U:51.722 V:51.930 Avg:50.043 Global:49.971 kb/s:703.44

encoded 1002 frames, 11.82 fps, 703.60 kb/s

build 999 command line
C:\Documents and Settings\SHiSH\Desktop\x264>x264.Build.999.exe --crf 14.0 --ref 16 --mixed-refs --no-fast-psk
ip --bframes 16 --b-adapt 2 --b-pyramid --weightb --direct auto --subme 9 --trellis 1 --partitions all --8x8d
ct --me umh --merange 32 --threads auto --thread-input --progress --no-dct-decimate --output "I:\Work\higurash
i\dvd5\dvd2\s21.build999.mkv" "I:\Work\higurashi\dvd5\dvd2\VIDEO_TS\s21.avs"
avis [info]: 848x480 @ 23.98 fps (1002 frames)
x264 [info]: using cpu capabilities: MMX2 SSE2Fast SSSE3 PHADD SSE4 Cache64
x264 [info]: slice I:7 Avg QP:10.56 size: 56936 PSNR Mean Y:52.06 U:54.48 V:54.44 Avg:52.72 Global:52.69
x264 [info]: slice P:702 Avg QP:10.70 size: 12962 PSNR Mean Y:50.73 U:53.26 V:53.31 Avg:51.43 Global:51.40
x264 [info]: slice B:293 Avg QP:11.78 size: 2044 PSNR Mean Y:50.12 U:52.67 V:52.68 Avg:50.82 Global:50.69
x264 [info]: consecutive B-frames: 52.0% 20.1% 19.0% 8.4% 0.5% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 0.0%
x264 [info]: mb I I16..4: 9.2% 76.4% 14.5%
x264 [info]: mb P I16..4: 0.2% 0.5% 0.1% P16..4: 65.1% 10.7% 21.7% 0.1% 0.1% skip: 1.7%
x264 [info]: mb B I16..4: 0.0% 0.0% 0.0% B16..8: 27.8% 0.5% 2.1% direct: 1.8% skip:67.8% L0:36.8% L1:55.4% BI: 7.8%
x264 [info]: 8x8 transform intra:71.9% inter:38.3%
x264 [info]: direct mvs spatial:87.7% temporal:12.3%
x264 [info]: ref P L0 80.7% 8.6% 3.7% 1.2% 0.9% 0.9% 0.8% 0.4% 0.4% 0.4% 0.4% 0.4% 0.3% 0.4% 0.3% 0.2%
x264 [info]: ref B L0 76.5% 16.7% 3.1% 0.9% 0.6% 0.9% 0.4% 0.2% 0.2% 0.1% 0.1% 0.1% 0.1% 0.1% 0.1%
x264 [info]: ref B L1 94.5% 5.5%
x264 [info]: SSIM Mean Y:0.9937798
x264 [info]: PSNR Mean Y:50.565 U:53.097 V:53.131 Avg:51.259 Global:51.190 kb/s:1932.77

encoded 1002 frames, 7.41 fps, 1932.93 kb/s

I want use latest build, but with waste of bitrate its imposible
Any advice?

[Sagekilla]

Please read the updates.. there's been a recent change that makes CRF allocate bits differently. You'll have to adjust your CRF to a higher value now to get the same bitrate you used to get. But you should get more or less around the same quality.

[XPecto]

Where i should look for crf Changes?
in x.264 help or is there any help around here?
Can you please post link. Thanks.

[Dark Shikari]

I made a major change to ratecontrol that changed the way CRF worked. Unfortunately, despite the fact that I tried to keep the bitrate the same overall, some sources' bitrate rose a lot, while others dropped, so the only thing I could do is try to "center" things overall.

As you might notice, CRF 14 is now ridiculous overkill; you don't need such a low value for any sane purpose. IMO the new scale is much much better; there probably no longer exist sources for which one "needs" CRF14 to achieve transparency, for example.

[lexor]

As you might notice, CRF 14 is now ridiculous overkill; you don't need such a low value for any sane purpose.
Actually, you might. Consider my use case. I would rip an anime and re-encode for both my stand alone and PS3 (ends up at about 2gigs, usually a bit less). Now I also have a PSP, and I found it to be much faster (in terms of encoding time) to re-encode that 2gig rip, than getting the source out again (which is usually blu-ray, so no mean feat to have a lot of those at a time on hdd).

In this case ~crf13 was perfect for me, producing about 150mb of video. Now obviously it is higher now, but my point is that you do have legit uses for low crf (in older builds).

[Sagittaire]

I made a major change to ratecontrol that changed the way CRF worked. Unfortunately, despite the fact that I tried to keep the bitrate the same overall, some sources' bitrate rose a lot, while others dropped, so the only thing I could do is try to "center" things overall.

As you might notice, CRF 14 is now ridiculous overkill; you don't need such a low value for any sane purpose. IMO the new scale is much much better; there probably no longer exist sources for which one "needs" CRF14 to achieve transparency, for example.

Not more simple to use always quant at N in average for Pframe if you use crf N ?

[cogman]

Actually, you might. Consider my use case. I would rip an anime and re-encode for both my stand alone and PS3 (ends up at about 2gigs, usually a bit less). Now I also have a PSP, and I found it to be much faster (in terms of encoding time) to re-encode that 2gig rip, than getting the source out again (which is usually blu-ray, so no mean feat to have a lot of those at a time on hdd).

In this case ~crf13 was perfect for me, producing about 150mb of video. Now obviously it is higher now, but my point is that you do have legit uses for low crf (in older builds).

What do you gain by using CRF 14? CRF 18 will look visually the same as the input for 99.99% of people. Even on a second generation encode you won't see any blocking that wasn't already present in the orgional.

I could see a potential case for 17 in this situation, but 14 and 13 are ridiculous, they aren't gaining you anything other then big files.

[ajp_anton]

Actually, you might. Consider my use case. I would rip an anime and re-encode for both my stand alone and PS3 (ends up at about 2gigs, usually a bit less). Now I also have a PSP, and I found it to be much faster (in terms of encoding time) to re-encode that 2gig rip, than getting the source out again (which is usually blu-ray, so no mean feat to have a lot of those at a time on hdd).

In this case ~crf13 was perfect for me, producing about 150mb of video. Now obviously it is higher now, but my point is that you do have legit uses for low crf (in older builds).Of course it depends on what you're doing with your encode. You didn't think DS meant we should all drop lossless encoding because x264's crf=18 is perfect for everything?

[lexor]

You didn't think DS meant we should all drop lossless encoding because x264's crf=18 is perfect for everything?

Huh? Who claimed that? I don't think anyone claimed crf18 was perfect for everything, it's always been merely a guide to start searching for high quality.

DS made a blanket statement that there is no sane use for low crf, I merely pointed out that there is. Nothing more.

[kemuri-_9]

lossless compression exists for your supposed reason...
low crf's as visually transparant as they are, still don't retain all the information like a lossless.
And then since you're encoding from an h.264 source file, you now have to worry about frame determinism, based on the decoder you're using and how you use it to make a new re-encode.
all in all, people use losslesses for the reason you state, not low crf encodes

[lexor]

People most definitely don't use Lossless for the reason I state, because PSP doesn't do lossless. The reason I used crf so low is because any higher in older builds used to crush the bitrate for some reason (in this use case) and produce poor results (I always assumed it was a price of re-encoding from relatively low bitrate source to even lower bitrate).

[kemuri-_9]

what? you appeared to have greatly misunderstood what i was saying at all.
I wasn't stating that people use lossless on PSP, i was stating that people use lossless files when they have a necessity to encode a single file to multiple formats.

it's the concept of taking and filtering the source into a single lossless, then encoding that lossless to all the formats desired.
so for your case it would look something like:

/-> PSP
source -> lossless -> standalone
\-> PS3

all the lossy encodes take from the lossless, not from re-encoding one lossy encode into another.
and because you're encoding from the lossless, there's no quality loss from doing such a lossy -> lossy re-encode

[STaRGaZeR]

What do you gain by using CRF 14? CRF 18 will look visually the same as the input for 99.99% of people.

I need CRF 16-16.5 to consider it transparent, depending of the source.

[lexor]

Last I tested, Lossless was frequently larger than blu-ray source. So, no, I don't see advantage of that middle man. Just use source, the whole point of using lossy encode is to speed up encoding. That certainly won't happen with a lossless source, it's just as big image dimension wise and probably even larger in bitrate. If anything it will slow things down.

[Sagekilla]

Low CRF values are still handy for extremely minimal lossy encoding. I used CRF 12 for encoding a particularly difficult sequence (720p60) and managed to cut the bitrate by 1/3 (source was ... 100 mbps lagarith?). It was enough bitrate to make sure the picture didn't suffer too much, but there were still some changes because of the inherent complexity of my source.


Either way, we have CRF 0 to 51 to give the luser enough rope to hang himself :)

[kemuri-_9]

Last I tested, Lossless was frequently larger than blu-ray source. So, no, I don't see advantage of that middle man. Just use source, the whole point of using lossy encode is to speed up encoding. That certainly won't happen with a lossless source, it's just as big image dimension wise and probably even larger in bitrate. If anything it will slow things down.

if you're going from source -> lossless without filtering, then there is no point/need for a lossless, just encode all the pieces from the source.
lossless is there to save the filtering done without data loss to a lossy encode.
so if there's no to very minimal filtering, then there is no point in having a lossless.

[Kurth]

I have been encoding my animes using --crf 18 --qpmin 18 and the quality is very good to my eyes (I watch anime on a LCD monitor with DVI) and the final file size is very good. (640x480 and 704x400 at 80MB~140MB depends on each anime episode)
Some times I add a weak deblock and weak denoise and it help some more too and don't destroy the image.
Im always encoding using the latest x264 build.
Oh and turn off the Psy-RDO this thing like to fuck the image and double the amount of bitrate used.

[Sagekilla]

@Kurd PSY-RD works better on real life imagery. It's better to turn it off for anime and then use it at a small strength (0.5:0.5 to 0.7:0.7) on films. If you adjust CRF accordingly (or if use 2-pass) too, using PSY-RD produces a better result than without.