Alternative to compressiblity check with x264

[digidragon]

Reading some posts here it seems that a compressiblity check with x264 isn't very meaningful. I'm used to doing a check before encoding with XviD, so what's the alternative way when encoding in x264 when using, say, MeGUI?

For example, using XviD with GK, I'd adjust the bitrate or resolution so that the bits/pixel*frame ratio is around 0.17 (to get an acceptable quality for a given file size). But is there another method I can use with x264, or do I just need to do it the old-fashioned way by doing test encodes and looking at the results...?

Thanks

[stax76]

Reading some posts here it seems that a compressiblity check with x264 isn't very meaningful.


Does those posts incl. technical background information, I'm interested to learn about this topic.


For example, using XviD with GK, I'd adjust the bitrate or resolution so that the bits/pixel*frame ratio is around 0.17 (to get an acceptable quality for a given file size).


bits/pixel*frame aka BPF isn't very meaningful either since movies vary greatly in compressibility, that's why StaxRip doesn't even show this value.

[Sharktooth]

Compressibility check with h.264 is a bit more complicated than usual coz there are more factors that may influence the result.
inloop deblocking filter is one of those...
There is (i think) no way to accurately calculate compressibility without compressing a very large part of the movie, and that makes the compressibility check useless coz of the time needed to do it.

[digidragon]

No, there's was no technical reason, I just got the feeling from what I'd read that it wasn't that meaningful with x264.

[bond]

iirc there are technical reasons why comp test for avc doesnt make much sense, but i dont remember why :D

anyways this has surely been discussed already so :search: for finding out the reasons

[Selur]

iirc cabac might behave not the same on a small collected part of the movie compared to compressing the whole thing,...

[Manao]

Actually, compressibility tests are as useful for ASP than for AVC. You just have to choose the quantizer that give for you the best needed quality ( it'll play the role of quantizer 2 in XviD ), and then to change the compressibility ratio you were used to aim at.

But anyway, I never found compressibility tests usefull ( once you know how the codec behave, you can predict roughly what bitrate you'll need to aim. Furthermore, back in the time when I was doing ripp, the choice was 1 or 2 CDs, the gap between the two being wide enough for lucky guess to always be lucky ).

[CruNcher]

@Manao
:goodpost:

[Sharktooth]

Im not convinced at all since i tried several compressibility test with different methods and they all failed to give me accurate info...

[buzzqw]

well i usually do a compression test is this way

encode a 5% using selectrangevery(300,15) at fixed quant of 22

then apply

FinalWidth=16*original_width*(((movie_size_in_byte_wanted-size_of_audio_in_byte)/0.99)/(sample_movie_5%_size_in_byte*13))

movie_size_in_byte_wanted is like 700mb*1024*1024
size_of_audio_in_byte is the audio size already encoded

total empiric solution but for me is ok (this is the formula used in my x264 gui MkvMagic)

BHH

[Episodio1]

When encoding to XviD there is a window showing info about frames and quants. When averaged quant is below 3'5 it is supposed to be a very good rip.

Is there a desired quant for AVC ? Is it more useful to use bitsperpixel instead?

Thanks.

[Audionut]

Is there a desired quant for AVC ? Is it more useful to use bitsperpixel instead?

I think your best bet is to encode the same clip at several different quants. Say 18 through to 26, and see what suits your needs best.

ASP quant 2 = approx AVC quant 18.

However for my eyes atleast, AVC quant 18 looks much better that ASP quant 2.

Test some, then test some more.

[foxyshadis]

x264 also has a 1-pass quality based mode, --qrf, which is not the "constant quant" quality of ASP codecs but actually attemps to code with a constant metric, I believe psnr. (Take that as you will.) It should give you an idea of the quantizer average you need and the output filesize of your desired quality. Not so useful if you're constrained by size, though.