x265 HEVC Encoder

[filler56789]

x265.exe 3.2+17-4a29e0c5bfaf

Download: http://www.mediafire.com/file/ieswxu...5bfaf.rar/file

FWIW...

In file included from D:/KOMPILES/MCW/x265/source/encoder/encoder.cpp:27:
D:/KOMPILES/MCW/x265/source/encoder/encoder.cpp: In member function 'void x265::Encoder::destroy()':
D:/KOMPILES/MCW/x265/source/common/common.h:222:37: warning: macro expands to multiple statements [-Wmultistatement-macros]
  222 | #define X265_FREE_ZERO(ptr)         x265_free(ptr); (ptr) = NULL
      |                                     ^~~~~~~~~
D:/KOMPILES/MCW/x265/source/encoder/encoder.cpp:882:12: note: in expansion of macro 'X265_FREE_ZERO'
  882 |            X265_FREE_ZERO(m_edgePic);
      |            ^~~~~~~~~~~~~~
D:/KOMPILES/MCW/x265/source/encoder/encoder.cpp:881:9: note: some parts of macro expansion are not guarded by this 'if' clause
  881 |         if(m_edgePic != NULL)
      |         ^~

[Selur]

found an issue about 2pass crashing (https://bitbucket.org/multicoreware/...ure-with-32-15)
->
scenecut-aware-qp=0 (default) fails
scenecut-aware-qp=1 works

hopefully they will fix this soon,....

[LigH]

@filler56789: Known and about to be fixed (it's really just a question of C source formatting).

[Selur]

'scenecut-aware-qp'-bug is fixed

[filler56789]

x265.exe 3.2+18-f0fe46ce379d

(64-bit, multilib, GCC 9.2.0)

http://www.mediafire.com/file/bqmxz9...e379d.rar/file

[Boulder]

Some new stuff there I see, has anyone made any tests? Once again there are no use cases available from the devs.

[filler56789]

Pooja Venkatesan committed a4e060a

2019-09-11

Add option hme-range to modify search range for HME levels L0, L1 and L2.

https://bitbucket.org/multicoreware/...49f3ba8b63ac5a

Download the .EXE from: http://msystem.waw.pl/x265/

[_kermit]

those are my current parameters for 1080p encoding and I'll use between CRF 18 to 22, depending on the source material

--preset slow --vbv-maxrate 20000 --deblock=-6:-6 --no-strong-intra-smoothing --amp --aq-strength 1.25 --qcomp 0.7 --no-deblock --no-sao --rskip --output-depth 10

in general I get good results but once in a while there are artifacts, particular in dark areas, looking like this:
(in this case it was CRF 22)



two questions:

any suggestions on improving my parameters in general?

Is there a parameter which would improve in particular on that (besides a lower CRF value of course)?

cheers, roland

[Boulder]

--aq-mode 3 is meant to battle issues with dark areas of the frame. You may want to tune the strength down though, the bitrate will most likely shoot through the roof

[markiemarcus]

Originally Posted by _kermit

those are my current parameters for 1080p encoding and I'll use between CRF 18 to 22, depending on the source material

--preset slow --vbv-maxrate 20000 --deblock=-6:-6 --no-strong-intra-smoothing --amp --aq-strength 1.25 --qcomp 0.7 --no-deblock --no-sao --rskip --output-depth 10

in general I get good results but once in a while there are artifacts, particular in dark areas, looking like this:
(in this case it was CRF 22)

Attachment 17045

two questions:

any suggestions on improving my parameters in general?

Is there a parameter which would improve in particular on that (besides a lower CRF value of course)?

cheers, roland

As mentioned above, I'd also look at Aq mode 3 with the default AQ strength. You may even be able to get better results (at the same file size) with around 0.85, even lower if the source is very grainy.

Additionally, I have yet to see any real detail benefit to no-strong-intra-smoothing. I'd also avoid disabling deblocking and, if grain retention is still a priority, use some light deblocking instead. I've had good results with -4: -1 and -3: -3, with the former being slightly better IMO.

--selective-sao 1 also seems to work well from my testing, subjectively better than disabling it entirely IMO.

[_kermit]

Originally Posted by markiemarcus

As mentioned above, I'd also look at Aq mode 3 with the default AQ strength. You may even be able to get better results (at the same file size) with around 0.85, even lower if the source is very grainy.

Additionally, I have yet to see any real detail benefit to no-strong-intra-smoothing. I'd also avoid disabling deblocking and, if grain retention is still a priority, use some light deblocking instead. I've had good results with -4: -1 and -3: -3, with the former being slightly better IMO.

--selective-sao 1 also seems to work well from my testing, subjectively better than disabling it entirely IMO.

thanks both.

I've gathered those parameters from many sources, and honestly have mostly no clue about the impact.

Something like this?

--preset slow --vbv-maxrate 20000 --deblock=-4:-1 --amp --aq-strength 0.85 --qcomp 0.7 --selective-sao 1 --rskip --output-depth 10 --Aq-mode 3

I don't care about grain. Grain is IMO just noise.
And while I understand that old material just has to have grain I don't understand why there is grain in newer material.

If grain can be reduced without having other negative effects on a movie, that would also be fine.
I don't analyze still pics, I don't see a point there. I'm not watching still pics.

would the above be a good compromise, sufficient, or is there more to fine tune?

-roland

[markiemarcus]

Originally Posted by _kermit

thanks both.

I've gathered those parameters from many sources, and honestly have mostly no clue about the impact.

Something like this?

--preset slow --vbv-maxrate 20000 --deblock=-4:-1 --amp --aq-strength 0.85 --qcomp 0.7 --selective-sao 1 --rskip --output-depth 10 --Aq-mode 3

I don't care about grain. Grain is IMO just noise.
And while I understand that old material just has to have grain I don't understand why there is grain in newer material.

If grain can be reduced without having other negative effects on a movie, that would also be fine.
I don't analyze still pics, I don't see a point there. I'm not watching still pics.

would the above be a good compromise, sufficient, or is there more to fine tune?

-roland

It's very difficult to compare without 2-pass, but I'd also ditch qcomp 0.7 . Between that and the Aq strength changes, bitrate at CRF 22 is going to drop quite a lot, so I'd maybe look at using a lower CRF. Maybe start around CRF 20.5 and see how you get on?

Perhaps keep AQ strength at the default of 1.0 for the time being. You really need 2-pass and short scene tests to find the optimal value.

[Boulder]

Losing grain means also losing details, but --selective-sao 1 probably helps with that. SAO used to be one option to disable (I still disable it completely myself), because it just smoothed everything. Otherwise, your settings should be a good basis. I don't know much about the inner workings of the deblocking feature, but I use --deblock -2:-2 myself.

[markiemarcus]

Originally Posted by Boulder

Losing grain means also losing details, but --selective-sao 1 probably helps with that. SAO used to be one option to disable (I still disable it completely myself), because it just smoothed everything. Otherwise, your settings should be a good basis. I don't know much about the inner workings of the deblocking feature, but I use --deblock -2:-2 myself.

Deblock -2:-2 usually looks nice from my testing also. I'd definitely give --selective-sao 1 a try. Helps with a few rough edges. You can really see the grain smearing start to kick in with --selective-sao 2 though.

[_kermit]

Originally Posted by markiemarcus

Deblock -2:-2 usually looks nice from my testing also. I'd definitely give --selective-sao 1 a try. Helps with a few rough edges. You can really see the grain smearing start to kick in with --selective-sao 2 though.

I've tried this on a 1 min. clip:

--vbv-maxrate 15000 --deblock=-2:-2 --amp --aq-strength 0.85 --selective-sao 1 --rskip --output-depth 10 --aq-mode 3

with CRF 18-22 in 1.0 steps.

CRF 18 results in 5.1Mbit/s (the original has 37Mbit), but all still have the same issue.

Anything else that can help?
Obviously otherwise only using very high bitrate might solve this or not re-encoing at all or accepting that this might happen.

[markiemarcus]

Originally Posted by _kermit

I've tried this on a 1 min. clip:

--vbv-maxrate 15000 --deblock=-2:-2 --amp --aq-strength 0.85 --selective-sao 1 --rskip --output-depth 10 --aq-mode 3

with CRF 18-22 in 1.0 steps.

CRF 18 results in 5.1Mbit/s (the original has 37Mbit), but all still have the same issue.

Anything else that can help?
Obviously otherwise only using very high bitrate might solve this or not re-encoing at all or accepting that this might happen.

Perhaps stick with the default Aq strength for the time being. CRF 18 with Aq mode 3 really should look good.

I take it you're running the latest build?

In the interest of narrowing the problem down, maybe try the "slower" preset. I'd also see if --qg-size 64 can help and there's always good old Aq-mode 1.

Is the source very grainy?

[Boulder]

Have you checked the original closely so that it doesn't have the same issue there? Re-encoding could easily amplify it.
It could be worth a try applying an anti-banding filter just to see if it changes the situation.

[_kermit]

Originally Posted by markiemarcus

Perhaps stick with the default Aq strength for the time being. CRF 18 with Aq mode 3 really should look good.

I take it you're running the latest build?

In the interest of narrowing the problem down, maybe try the "slower" preset. I'd also see if --qg-size 64 can help and there's always good old Aq-mode 1.

Is the source very grainy?

Tried now:

--vbv-maxrate 15000 --deblock=-2:-2 --amp --selective-sao 1 --rskip --output-depth 10 --aq-mode 3 --qg-size 64 --preset slower --crf 20

and

--vbv-maxrate 15000 --deblock=-2:-2 --amp --selective-sao 1 --rskip --output-depth 10 --aq-mode 1 --qg-size 64 --preset slower --crf 18

yes, 3.2.0.18

still shows. It may be less, but still obvious

Interesting part is that it only is that obvious on that region with black/purple texture

it is grainy, but there is worse.

does it help uploading the original and encoded clip?
Onedrive working?

[_kermit]

Originally Posted by Boulder

Have you checked the original closely so that it doesn't have the same issue there? Re-encoding could easily amplify it.
It could be worth a try applying an anti-banding filter just to see if it changes the situation.

the original doesn't have that.

I have no idea how to apply such a filter

[Boulder]

A source clip can always be useful. Easier to see what's really happening.