x.264 Target Quality

[Funkstar De Luxe]

Hi,

I'm going to start transcoding all my DVDs to x.264 with the help of Staxrip. One thing that confuses me here is the target quality setting. Since I am keeping them all on my HDD I don't need to hit a certain file size, I would like transparent encoding though.

My question is, is only 100% target quality transparent or can I go with something like 75%? The reason I ask is becuase anything over 70%, staxrip askes me to lower the bitrate.

Thanks!

Tony

PS, one last question I'll stick in here. For a standard stereo track should I encode to AAC LC VBR or AAC HE VBR?

[siddharthagandhi]

I'm not 100 percent sure, but I believe AAC LC VBR is better.

And why don't you specify the bitrate instead of using a target quality? I would suggest a bitrate of 8-10 mbps for hard drive storage for 100 percent transparency.

[Funkstar De Luxe]

I tried specific files sizes first, giving me bitrates of around 8-10mbps, but the quality was not too good.

I guess I am best using one of these profiles:
CQ-ASP_Q2_equiv: Almost the equivalent of MPEG-4 ASP Q2 encoding.
CQ-ASP_Q2_eq(crf): Same as above but with "constant quality" (constant rate factor).

However I do not know the difference between them. Can anyone tell me what one is best?

[nm]

Crf is better.

I have to wonder, why would you want to re-encode if 6 Mbps or more is fine as a resulting bitrate? Just rip the DVDs to DVD image files and be done with it :)

What was wrong with the 8 Mbps encodes, by the way? Smoothed texture in flat areas or something else?

[Funkstar De Luxe]

Ah, I am an idiot. I'm getting all confused here guys, can I start again? :-D

Here is what I'd like. A good quality copy of my DVDs stored on my PC. No need for specific bit rates or files sizes, just good quality.

I'd like to use x264 becuase I think it will be more future proof and provide the best quality to size ratio.

What profile do you suggest I use?

[siddharthagandhi]

now that i think of it, nm is exactly right

get dvd decrypter and rip the DVD to your hard drive

then you can convert the vob files to mpeg using MPEG Mediator

and what kind of movie is it that it still has quality problems even at a bitrate of 10 mbps?????

[Funkstar De Luxe]

Hold on, does 10mbps = 1000 kbps?

[nm]

Nope, 10 Mbps (http://en.wikipedia.org/wiki/Megabit_per_second) is 10000 kbps (could actually be 10240 Kibit/s in some cases).

[Funkstar De Luxe]

Ok. I'm sorry I've been such a noob here, I really appreciate all the time you guys took to answer my questions. I'm going to run some test and then I'll be back with more.

Thanks again!

[foxyshadis]

That's 1 mbps.

How many DVDs are we talking, and how much storage are you willing to dedicate? For average a gig a dvd, you need light filtering, slow compression, and crf ~24. For two gigs a dvd, minimal filtering, faster compression, and crf ~18 (like 1P-Intermediate, but crf). If you've got an enormous dedicated media center hard drive, well, go for crf 14-16 with no filtering and minimal deblock settings.

[siddharthagandhi]

He wants 100% transparency. Thats talking pretty big hard drive space.

[foxyshadis]

sidd, how do nero % translate into quant/crf values, do you know? I never used that mode in recode, and I don't have it on here ATM to test, but it'd be really nice to know.

[stax76]

It's a difficult topic I'm not very good in since I still use XviD as my CPU is too slow for x264, values between 40-60% quality for multipass and crf 20-24 for single pass are probably good. You can customize in StaxRip's codec configuration how much quality StaxRip should work with. You can just click next skipping the warning about quality btw.

[Deinorius]

With Constant Quality I'm using crf 24. Quality is quite good and Filesizes are small. For better quality you can use 22 too, but I wouldn't use higher crf values.

For a standard stereo track I would use Vorbis aoTuV 4.51 (Quality 0-2) or HE-AAC at low bitrates but I prefer Vorbis for Stereo or DLPII.

[Funkstar De Luxe]

@Stax: I'll give it a go at 22. Problem is that it encodes unbelievably slow. 2.15 FPS on my 3.6Ghz. So about 15 hours per movie. Thanks a lot for the great tool, I'm a total newbie at video coding but Staxrip let me use the best technology very easily. Much respect!

@Deinorius: I wasn't aware that the mp4 container would let me use vorbis. I'll certainly give it a try, I'm much more confortable with audio codecs than video.

Also, can you tell me how long it takes you to encode a film in cfr? I think I may be missing some settings becuase 15hours seems like a damn long time. All I did was load up the cfr profile in Staxrip and change the quality to 22. I suppose if it give me the quality I am looking for I can't complain...

Tony

[Deinorius]

Which profile are you using? I have only an Athlon XP-M 2000+ and therefore I can only encode with faster settings.
I normally encode Animes and I'm using Sharktooths profile "AE_Standard" with M.E. Algorithm - Hexagonal not Multi-hexagonal. It's fast and good enough for bitrates around 500 kbit/s and lower(640x480).
You can use slower settings like multi-hex or subme 6, which gives you better quality. Try CE_MainProfile.

If you want to use the Rips only on your HDD, then you can use Matroska too. You can use Vorbis and AAC and SSA Subs. It's your decision.

[foxyshadis]

You don't have to use the slowest settings unless you're trying to squeeze every byte out of the file. Faster settings would give you the same quality in a larger file. (Filtering might cause some of the slowdown though.) I'm not too familiar with staxrip's profiles, but I'm sure stax can help you choose a faster one.

Also, mp4 doesn't support vorbis, that's for mkv; only aac is really mp4-compatible.

[stax76]

StaxRip uses Sharktooth's profiles, for general purpose encoding there is 'CE-Highprofile', CE means common encoding.

[siddharthagandhi]

"sidd, how do nero % translate into quant/crf values, do you know? I never used that mode in recode, and I don't have it on here ATM to test, but it'd be really nice to know."

They dont use percentage. You can only specify the bitrate.

You're asking about target quality percentage right?

[cbarroso]

So, is the following overkill?

--crf 18 --ref 8 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --b-rdo --bime --weightb --filter -2,-1 --subme 6 --trellis 2 --analyse all --8x8dct --me umh --progress --output

[foxyshadis]

At crf 18, probably, unless you're really hurting for space and need every bit. In particular, trellis 2 and b-rdo are speedkillers, and you could probably get away with subme 5 (which also shuts off the other two). They're mostly effective when you have a clean or filtered video; if it's noisy rdo loses much of its effectiveness.

siddhartha, yeah, I meant target quality %.

[starg4ze]

Nope, 10 Mbps (http://en.wikipedia.org/wiki/Megabit_per_second) is 10000 kbps (could actually be 10240 Kibit/s in some cases).

To be correct(which of course is the ultimate goal for all of us...) it's like this:

1 Mbit is 1 000 000 bits/s e.g 1000 kbit/s. Period.

If someone tells you 1024000 bits/s is 1Mbit/s, he's wrong.

When it comes to Bytes...now thats another story...here we have 3 different definitions of 1MB:
From Wikipedia: (http://en.wikipedia.org/wiki/Megabyte)

1. 1,048,576 bytes (1,0242, 220): This definition is used for nearly all discussions of computer memory (as computer addresses are naturally powers of two, making it efficient to manufacture memory in power-of-two capacities) and file storage. As of 2005, most software uses this definition to express storage capacity (e.g. file size).
2. 1,000,000 bytes (1,0002, 106): This is the definition recommended by IEC. It is used primarily in networking contexts and most storage media, particularly hard drives and DVDs. This definition of 'mega-' as a "binary prefix is consistent with the other SI prefixes, and with many other uses of the prefix in computing, such as CPU clock speeds or measures of performance.
3. 1,024,000 bytes (1,024×1,000): This definition occurs rarely. It was used in a small number of storage contexts, most notably the "1.44 MB" (actually 1,474,560 bytes) "3.5-inch" (actually 90 mm) high-density floppy diskette.

[nm]

To be correct(which of course is the ultimate goal for all of us...) it's like this:

1 Mbit is 1 000 000 bits/s e.g 1000 kbit/s. Period.
I know what you mean: in telecommunications, multimedia and most other places this is the common meaning. I was just pointing that there are rare exceptions where 2^n units are used instead, and then the correct prefix would be Mibit/s, which is unfortunately not always used when needed.

[OvejaNegra]

i always encode in CQuantizer mode, i like q and the files are very small.
if you use constant Quantizer all the frames will receive the same quantizer (the i frames will be boosted and the Bframes will be lowed, that´s normal) the visual quality will be the same for EVERY frame. Fast action or low action, the same q for both. Try a quantizer of 24 or less (22 - 20 - 18 are my favorites) 22 is ok for animes, for movies i prefer 20 or 18 for a masterpiece like JPark. If you use constant quality, fas action escenes will receive a lower quantizer (you won´t notice) but sometimes YES. Try constant quantizer with 22-20 or 18.
And you must adjust other settings for speed.
ALL my encodes are constant quantizer.
It´s a VERY GOOD choice.
Please Try.


Sorry for my english
Good Luck:
If :you have a HP Server with 4 CPUs try
x264.exe --qp 22 --ref 5 --mixed-refs --no-fast-pskip --bframes 2 --b-rdo --bime --weightb --nf --subme 7 --analyse all --8x8dct --me esa --progress --no-psnr --output "" "" (with 4 threads enabled ;-)


else


try this


x264.exe --qp 22 --ref 3 --mixed-refs --no-fast-pskip --bframes 2 --b-rdo --bime --weightb --nf --subme 6 --analyse all --8x8dct --me umh --progress --no-psnr --output "" ""

end sub ;-)

again sorry for my english

[Manao]

the visual quality will be the same for EVERY frameNo. I could show you videos encoded at quantizer 36 that looks good, and other that are plainly unwatchable. What happens in your case is that Q20 is so good in quality that you can't perceive the visual quality variation. But it's there, and it would have been seen if the quantizer was greater.

That also implies that if you let the quantizer vary slightly, you wouldn't perceive a visual quality variation. Which means that in your case, crf or 2passes encoding would yield the same constant quality, but at a smaller bitrate due to a saner bitrate usage

[OvejaNegra]

that´s interesting
i have questions:
how can i avoid the quantizer lowering in the high speed scenes?
when you say you have videos with a q of 36, what kind of source?
it´s clean? fast action? low action?


sorry for my english again
thanks

BTW: when i say: the visual quality will be the same for EVERY frame,
i meant that all the frames will have the same quantizer (except the If and the Bf) and the visual quality will be the same of that quantizer.
wrong?

[Manao]

It was a high quality, noiseless video of a public watching a whale jump out of the water. Not quite fast action but it gives insanely high bitrates, even at quantizer 36.

When encoding without targetting a quantizer, the quantizer will raise during fast action / hard to encode scenes. You can't prevent that behavior since raising it during these scenes allow huge improvements during all the other scenes ( provided that the bitrate stays the same ).

However, you can fix an arbitrary limit to the maximum quantizer with --qpmax.