X264 vs x265 at sane values

[Valnar]

OK, forgive some obvious n00b questions, but I'm only starting to dip my toe into H.265 after years of H.264, and MPEG2 before that.

I see a lot of comparisons between x264 and x265 (for instance) where they use extreme values for the encoding of either. Comparisons exist where they use the exact same bit rate (where x265 will be better of course) or use the same CRF, which is not fair since the default for x264 is 23 and x265 is 28.

What I want to know is....once all this testing is done and the product matures, what are the sane values going to be and what kind of results can we expect? Meaning.... I assume the happy medium will be, in order to make HEVC worth it, some level of bitrate reduction and some increase in quality. I'm pretty happy with x264 quality, so if an encode was a little better AND saved me some space, I'm all for that. What I see now, in most comparisons, is not quite realistic.

I know this is a completely subjective question, but lets say a 2 hour movie encoded with slow settings on x264, CRF 21 comes out to 4.5GB. What kind of space savings could I expect with similar settings on x265, and how much better will the video be to boot? Sure, I could try it myself, but I don't know what those "sane" settings are for x265 yet!

[jagabo]

I don't know what those "sane" settings are for x265 yet!

Neither does anyone else as the code is still in flux.

[sneaker]

Comparisons exist where they use the exact same bit rate (where x265 will be better of course)

Then you have not seen a lot of comparisons. x265 is often worse than x264 at the same bitrate. It all depends on source type and settings and with x265 being a young project no one will really be able to give you an exact answer at the moment.

or use the same CRF, which is not fair since the default for x264 is 23 and x265 is 28.

The defaults have nothing to do with it. It is unfair because simply e.g. --crf 20 will result in different qualities between x264 and x265. crf isn't even constant quality within a single encoder itself if you start adjusting settings. That's why proper comparisons always use the same bitrate.

[Valnar]

Thanks sneaker.

The point of my post is.... Lets assume the same bitrate makes x265 look better since its more efficient. Lets assume that to get the same video quality as x264, one can use less bits instead. Somewhere in the middle is where we will end up. I'd assume (yes...another assumption!) that the point of a newer codec is to increase the quality as well as reduce the size, and there is some fair balance there.

If we aren't there yet, I understand. Maybe this question is best asked a year from now. And by then, us mere mortals won't need to really know. Those decisions, arrived subjectively through a lot of testing by the programmers, will be incorporated into the next Handbrake or similar kind of app. I was just hoping to get an idea of what to expect!

[jagabo]

I'd assume (yes...another assumption!) that the point of a newer codec is to increase the quality as well as reduce the size

No. x264 and many other codecs can already give you 100 percent lossless encoding. It's all about reducing the bitrate. Whether you choose to use the same bitrate and get higher quality, or use a lower bitrate to get the same quality, or something in between is up to you. And don't forget encoding time. Not everyone wants to wait forever for their videos to encode.

[pandy]

Purpose of h.265 (x265) is to be twice efficient as h.264 (x264) - so or with same bitrate twice better (highly questionable as purely subjective) or with half bitrate same quality (and this is less disputable).
There is no sense to do comparison with same bitrate.

[Selur]

There is no sense to do comparison with same bitrate.

Why not ? It allows to use different quality metrics when comparing results.
If you choose a fixed quality metric and encode with x264 and x265 to compare the output file size you are stuck with that metric.
-> using a fixed file size (bitrate) seems to be the better choice for comparisons.

[Valnar]

No. x264 and many other codecs can already give you 100 percent lossless encoding. It's all about reducing the bitrate.

I don't think anyone normally encodes with x264 lossless, but maybe they do and I'm an outlier. That being said, even CRF 18 is not indistinguishable from a native HQ source and most people don't even use that. But if you can get CRF 20-22 with all the appropriate bells & whistles parameters (or whatever the equivalent is in x265 land) to provide lower file size AND slightly better quality, I'd say that was a win-win.

If anyone has comparisons that show that, I'd love to see it. Otherwise, I will wait until the codec matures.

[racer-x]

Tons of comparisons already done, do a search......

As of right now x264 beats x265 in everything except very low bitrates based on all my testing. My testing is the only ones I trust. You should do the same.

[pandy]

There is no sense to do comparison with same bitrate.

Why not ? It allows to use different quality metrics when comparing results.
If you choose a fixed quality metric and encode with x264 and x265 to compare the output file size you are stuck with that metric.
-> using a fixed file size (bitrate) seems to be the better choice for comparisons.

Because there is no objective way to perform subjective quality assessment, side to this as effect can give insignificant increase of quality and perfectly it can be related to metrics uncertainty.
Also because h.265 goal is bitrate reduction when compared to previous codec generation - keeping same bitrate sound odd.

[Selur]

Also because h.265 goal is bitrate reduction when compared to previous codec generation ...

Yes, one major goal was to produce a bit rate reduction compared to H.264, no argument there, but deducting that keeping the same bit rate when comparing H.265 against H.264 from that seems wrong and kind of leaves me puzzled.

From my point of view you either need to:
a. keep the output (~quality) the same and then compare the output size (bit rate)
or
b. keep the output size (bit rate) the same and then compare the output (~quality)
to decide if H.264 or H.265 is better under the given restrictions. (source, settings,...)

The benefit of using method b. over method a. is that the comparison can be done without the need to commit to a fixed quality metric. (Sadly no objective metric exists which reflects the human visual perception, PSNR, MSE, MSAD, SSIM, VQM all have their problems.)

[pandy]

Yes, one major goal was to produce a bit rate reduction compared to H.264, no argument there, but deducting that keeping the same bit rate when comparing H.265 against H.264 from that seems wrong and kind of leaves me puzzled.

From my point of view you either need to:
a. keep the output (~quality) the same and then compare the output size (bit rate)
or
b. keep the output size (bit rate) the same and then compare the output (~quality)
to decide if H.264 or H.265 is better under the given restrictions. (source, settings,...)

The benefit of using method b. over method a. is that the comparison can be done without the need to commit to a fixed quality metric. (Sadly no objective metric exists which reflects the human visual perception, PSNR, MSE, MSAD, SSIM, VQM all have their problems.)

As it is expected to have at least twice better quality but don't ask me about quantitative approach to quality - how to judge quality in quantitative way - have no clue for example that "leafs on trees" provide directly my brain to perceived twice better - strangely it can be asymptotic - twice better can be fraction from perceived improvement - it is like judging audio quality - at some point all codecs deal quite well and only trained listeners can perceive difference but still it will personal thing to judge on this "twice better".
That's why i found approach - same content, same quality but half bitrate more fair.

So once again = if you ask me: i don't know where is border between good and twice better than good - i found personally extremely difficult to answer such question (but i'm partially trained in video quality - i know where to search for distortions and see them sometimes in a way where i can't watch content but other people don't see any issue - even showing them with finger - "this is problem annoying me" they usually say they even not realize that this is problem...).
IMHO half bitrate is target for H.265 and it should be tested in such way (twice reduced when compared to H.264 and four times when compared to MPEG-2).

[ezcapper]

In regards to the original question, I think sane values for a casual comparison's sake using ffmpeg, x265 and x264 would be:

x265:
ffmpeg -i D:\1.avi -c:v libx265 -x265-params crf=23.0 -c:a copy the1.mp4
x264:
ffmpeg -i D:\1.avi -c:a copy -x264-params crf=23.0 the1.mp4

----
The input file is titled 1.avi and the output file is the1.mp4

For me at these values, the x265 is about half the bitrate of x264 and produces a slightly lower quality but still very watchable. The original input file was 4k live action roughly 16.x mbps. x265 is about 1/4 the speed of x264 in this type of test

It seems like they want to get h265 to the point where 5500 mbps or so 4k video (encoded using the crf parameter) looks excellent. They'll definitely get it within the next couple years. It already looks pretty good. Speed shouldn't be a problem as computers get faster and faster as long as the codec doesn't get slower.