x264 lossless and presets help

[reigan42]

Hi, i'm using footage from an hdpvr2 that has x264/aac in m2ts container and I was wondering does it negatively affect quality in lossless x264 to use ultrafast preset over slower?

[poisondeathray]

No - "lossless" means the quality is the same

The presets when used with lossless mode will affect compression ratio and speed (in general, slower presets will yield smaller filesizes)

Lossless will make your filesizes many times larger than the compressed original - unless you have a good reason or usage scenario to do that I would avoid it

[reigan42]

So i'm good to just use the ultra fast preset over slower. Thanks for the reply

[poisondeathray]

But why are you doing this ? It just makes the filesize larger, less compatible with various devices and software

It might be used if you were applying some filters or for intermediate use

[reigan42]

Well its footage from a ps2 so im gonna deinterlace, crop resize then finally use crf encoding for the final video. I just didn't want to lose any quality while using the filters.

[hello_hello]

Normally you'd de-interlace, crop, resize and re-encode in one step so there's no need for a lossless intermediate. Obviously sometimes there's reasons for using an intermediate lossless file (I sometimes do when using slow filtering if I'm being indecisive about any filtering that'd follow) but if it's not necessary it's probably just increasing processing time without benefit.

Or are you referring to an original lossless capture, in which case you can ignore my post entirely?

[reigan42]

Yeah im not for sure what sort of filtering path i want to do yet. So far using qtgmc to deinterlace and possibly spline64 for resize not sure on cropping whether i should use autocrop or regular crop have a hard time with mod2 restrictions on crop though. I'm re-encoding so the original file is a 14mbps x264 capture.

[hello_hello]

If you're resizing you can use the resize filters to crop and if you de-interlace first you can effectively crop mod1.

For example to crop 3 pixels from the left and three from the bottom while resizing to 1280x720 you could do it like this:

Crop(2,0,0,-2)
Spline36Resize(1280,720,1,0,0,-1)

[reigan42]

thanks for the tip i'll try that is it safe to deinterlace and crop in the same script as long as deinterlace is before the crop and resize?

[reigan42]

well i figured out my crop problem I originally did
crop(46,16,36,2)
instead of
crop(46,16,-36,-2) and it would give me the error resize filter source image too small.

[hello_hello]

thanks for the tip i'll try that is it safe to deinterlace and crop in the same script as long as deinterlace is before the crop and resize?

Yes.

[LMotlow]

The codec is h264, not "x264". x264 isn't a codec, it's an h264 encoder. BluRay/AVCHD has video structure and encoding requirements: http://forum.doom9.org/showthread.php?t=154533....unless, of course, you just don't care about the results that much.

[_Al_]

He did not mentioned Blu-Ray and Blu-Ray compatible settings are not better than other settings (non Blu-Ray specs), on the contrary, you get filesizes slightly bigger for the same CRF, because of short GOP, and some other limits, so compressibility (therefore quality if encoded to size) is in theory worse.

[LMotlow]

I agree. The thread's priority isn't quality or compatibility. I don't think the O.P.' understands the replies so far anyway, so the specs are moot.

[reigan42]

No, i get what you guys have been posting. I just wasn't sure on a few things. Yeah i could careless if my settings works on a bluray cause this is gonna be on youtube eventually.

Here's my final avs script if you guys are curious:

v=lwlibavvideosource(source="Video 1.m2ts")
a=lwlibavaudiosource(source="C:\Users\Robert\Video s\video 1.m2ts")
audiodub(v,a)
QTGMC( Preset="Medium", SourceMatch=2, TR2=2 )
SelectEven()
crop(46,16,-36,-2)
Spline64resize(640,464)
ConvertAudioTo16bit()

[addseo1115]

Very well, Thanks for sharing.

[usually_quiet]

The codec is h264, not "x264". x264 isn't a codec, it's an h264 encoder.

No, he's actually more correct than you are. People tend to use the word "codec" in informal conversation when it is more technically accurate to use the word "format" or "standard" instead. The formal definition of a codec is a device or software that encodes or decodes a digital data stream or signal. Don't believe me? Look it up in a good technical reference, or even Wikipedia.

h.264 is a video encoding format or standard. x264 encodes video according to that standard.

[sophisticles]

No - "lossless" means the quality is the same

The presets when used with lossless mode will affect compression ratio and speed (in general, slower presets will yield smaller filesizes)

Lossless will make your filesizes many times larger than the compressed original - unless you have a good reason or usage scenario to do that I would avoid it

This makes me wonder if x264's "lossless" mode is truly lossless. Consider this, x264's ultrafast preset disables CABAC, B frames, AQ, Psy-RD, Trellis, pretty much all of x264's options that are supposed to improve quality.

The more aggressive presets, such as very slow, enable and tweak these settings. Now here's the part that doesn't make any sense to me, things like AQ, MB-Tree and Psy-RD move bits around within a frame in an effort to improve perceived visual quality, i.e. by emphasizing portions, like a persons face, where a viewer is more likely to notice a quality difference at the expense of a background where the quality isn't as important (at least that's the official party line).

Now with a "lossless" encode, I don't see how it's possible to encode 2 versions of the same file, one with no Psy options where the bit distribution is even throughout the frame and one with Psy options enabled where the bits are distributed according to some algorithm and end up with 2 files that are mathematically lossless.

Now I assume you will counter with the fact that the one with the slower presets used less bit rate to achieve a lossless result but again since the algorithms used are different, to me, it just seems like it may not be as "lossless" as we are led to believe.

I personally wouldn't use x264 for "lossless" work and I really don't believe that too many pros do either.

[poisondeathray]

No - "lossless" means the quality is the same

The presets when used with lossless mode will affect compression ratio and speed (in general, slower presets will yield smaller filesizes)

Lossless will make your filesizes many times larger than the compressed original - unless you have a good reason or usage scenario to do that I would avoid it

This makes me wonder if x264's "lossless" mode is truly lossless. Consider this, x264's ultrafast preset disables CABAC, B frames, AQ, Psy-RD, Trellis, pretty much all of x264's options that are supposed to improve quality.


The more aggressive presets, such as very slow, enable and tweak these settings. Now here's the part that doesn't make any sense to me, things like AQ, MB-Tree and Psy-RD move bits around within a frame in an effort to improve perceived visual quality, i.e. by emphasizing portions, like a persons face, where a viewer is more likely to notice a quality difference at the expense of a background where the quality isn't as important (at least that's the official party line).

Now with a "lossless" encode, I don't see how it's possible to encode 2 versions of the same file, one with no Psy options where the bit distribution is even throughout the frame and one with Psy options enabled where the bits are distributed according to some algorithm and end up with 2 files that are mathematically lossless.

Now I assume you will counter with the fact that the one with the slower presets used less bit rate to achieve a lossless result but again since the algorithms used are different, to me, it just seems like it may not be as "lossless" as we are led to believe.

It is lossless, when used properly. This is a fact, and can be verified objectively

Psy options are automatically disabled in lossless mode

b-frames , CABAC, etc.. improve compression efficiency - which improves quality at a given bitrate. That concept of improving quality at a given bitrate only applies to lossy encoding, because by definition "lossless" is one defined state of quality

If you have lossless encoding, compression can still be improved by applying various techniques to improve lossless compression (e.g you can use temporal compression, b-frames etc.. instead of just I-frame only ), but the quality is the same eitherway, because "lossless quality" is the same quality at any bitrate. Lossless in the video compression context means decoded image is bit for bit identical. It does NOT mean compressed file is bit for bit identical. It's almost analgous to file compression algorithms. Some yield higher compression , some lower compression e.g. 7zip using lzma2 will often lead to better compression than say standard zip . There are filesize differences there. But when you decompress the archive the file is bit for bit identical, even metadata etc.. to the original. The difference in video however, is there is a decoding stage to uncompressed before the lossless compression is applied (you can't get back the original video with the original compression scheme, or metadata etc...)

You should read up on lossy vs. lossless encoding .

I personally wouldn't use x264 for "lossless" work and I really don't believe that too many pros do either.

If you mean for use in professional workflows, it's not used often. There is limited compatibility in "professional" programs (almost none) for lossless AVC