MeGUI Settings

[anirban]

What are the settings in MeGUI for best quality DVD to MKV conversion with x264 video?

[poisondeathray]

Absolute best quality would be x264 CQ-lossless profile, and original audio copied from your DVD source. But your file size is going to be huge.

[anirban]

Suppose I want to keep the video bitrate around 1800kbps, what would be the best settings.

[poisondeathray]

If you want a set average bitrate or filesize, use the HQ-Insane profile - it is preset for all settings maxed out. The only setting you change is the bitrate.

However, it takes about 2-3x longer to encode than the HQ-Slow profile and the difference in quality is negligible.

[PuzZLeR]

If you truly want "the best" with x264 in general, but don't want big files, use CRF=18 and one of the profiles with slower settings (as Poison explained). You can still use a faster setting with CRF, and still get similar quality, but your files will be bigger.

If you want the best possible quality at 1800kbps, then use bitrate encoding, again, for one of the slower settings to compress more quality into your file. You need to give the encoder more features (higher B-frames, reference frames, subme, bime, etc.) to grab that quality and it will need more time to provide it.

However keep in mind, some video will look much better at 1800kbps than others, and you won't know in advance. This is why I recommend CRF which adjusts the correct bitrate each time for you.

[Ethlred]

Best is a variable. Depends on your patience and the size of your monitor and just how finicky you are. Lossless is inherently the best result you can get. It also is slow and huge, usually you will get a larger file than the original. I doubt that you really want that. So now with best out of the way the question is time versus size versus quality.

So run tests. I recommend using the trailers for testing to find out what you like. Or you try adding this line to the AVS file.

SelectRangeEvery(600,12)

That way you can test out the whole movie in a reasonable time. The final result will be about 50 times as large so you can get a good idea of how much you compression you can use in comparison to the file size you want if you are using CRF instead of two pass.

I always use CRF which cuts the encoding time down to tolerable on my older PC. I start the encode at night before going to bed and when I get home from work its done. Sometimes even two movies can be done.

I use CRF settings from 18 to as high as 25. The high end is for stuff that is really hard to compress enough yet I can't justify the larger amount of space that the lower settings will produce. TV for instance. I just did the first season of Buck Rogers and even Pamela Helmsly isn't going to get a Gig an hour. I did most of them at CRF 21 after the first two episodes came out a bit larger than what I wanted. Wasn't so big I re-encoded, I did those two at CRF 20.

Settings I used

--crf 21.0 --level 4.1 --ref 3 --mixed-refs --no-fast-pskip --bframes 3 --b-rdo --bime --weightb --filter -1,-1 --subme 6 --trellis 2 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --threads auto --thread-input --progress --output "output" "input" --aq-mode 2 --aq-strength 0.7 --rdcmp psy

In the past I set subme to 4 as I found it was good enough for me and cut the encoding time considerably. However recently Dark Shikari has been testing a new patch for those like me that must have GRAIN if it was in the film to begin with. It requires a minimum subme setting of 6. I should test it at 7 to see if I can tolerate the time increase but haven't done so yet.

--aq-mode 2 --aq-strength 0.7 --rdcmp psy had to be set in the Zones tab in the Custom Command Line options space. Future versions of Megui may support one or both without using the Custom Command Line setting but its still experimental. However they aren't a problem for decoders so even if the experiments are dropped they will still play later.

AQ and Psy RDO just showed up in the latest version of X264 that Megui downloads automatically so you shouldn't need to hunt for the patched version on Doom9.

Psy RDO is discussed here:
http://forum.doom9.org/showthread.php?t=138293

It's replaced FGO as it is a more general purpose solution

AQ is intended to cut down on the blocking that often occurs in dark shadows or flat shaded areas. It may not be needed with Psy RDO. Many using CRF 18 are using to cut down on that sort of blocking so you may find CRF 20 to 22 enough.

--no-fast-pskip - When I encoded The Thirty-Nine Steps I found that x264 with pskip made it a blockfest. B&W brings that out. Turning it off fixed it. So I leave it that way.

--ref 3 Thats enough for me. More slows things down without much improvement in file size. More than 5 causes problems with hardware decoders. Which is why I just recently changed the level to 4.1. Didn't have to change anything else though so its just cosmetics for me. Wanted to be sure it would work if I get a hardware decoder.

--b-rdo instead of --b-pyramid because I often use Nero Showtime from Nero 6 which can't handle B frames used as references. Didn't make make much difference in file size so I have left it that way.

[Ethlred]

Just discovered that AQ is now on by default. I was running a test with it on in the command line and another with the it removed from the command line. Took longer with it off. Turned out to be on with higher settings by default.

--aq-mode <integer> How AQ distributes bits [2]
- 0: Disabled
- 1: Avoid moving bits between frames
- 2: Move bits between frames
--aq-strength <float> Reduces blocking and blurring in flat and
textured areas. [1.0]
- 0.5: weak AQ
- 1.5: strong AQ

So default was a little slower and the file is smaller. 24.4 MB @ 4.13 fps with strength of .7 versus 23.3 MB @ 3.57 fps.

--b-rdo RD based mode decision for B-frames. Requires subme 6.
--rdcmp Metric used for RD mode decision ["psy"]
- ssd: normal (maximum PSNR)
- psy: psychovisual (sharper)

Which means PsyRDO defaults to on IF you use subme=6 or 7. Otherwise its off, period. You can turn it off even with subme=6 or 7 by using the "--rdcmp ssd" setting

So if you ignore all this stuff and just leave the Megui Custom Command Line blank you will get

AQ mode 2 strength 1.0
PsyRDO on IF subme is 6 or 7.

Which means you can ignore all the crap I wrote about setting the Custom Command Line IF they keep the defaults in the future. I think I will set it explicitly.

The psy in PsyRDO stands for psychological. So it LOWERS the objective measurements like PSNR while making it look more detailed to the human eye. The higher AQ the default produced also lowered the PSNR which isn't surprising since it lowered the file size. I expected the opposite since I thought it would spend more bits to cut down on the blocking. Clearly it didn't work that way.