Interesting poll about codecs

[len0x]

On AutoGK's website there is a survey going on about which codecs ppl wanna see supported. I'm very suprised by number of ppl whishing for WMV9 support! (not that it matters coz its impossible anyway, but still very interesting) :)

[imf4]

What are the biggest things that make it impossible? I think I remember you mentioning it uses the registry or something?

I use it now to make files my Pocket PC and WMV-HD. I let AutoGK run through the first pass. Then I cancel and use the avs file. It works great for what I do.

[len0x]

Originally posted by imf4
I think I remember you mentioning it uses the registry?

no. it doesn't

[fewtch]

God, WMV9 looks so bad compared to other codecs...

Strange as it might sound, I'd like to see support for DivX v3.11 (because my crappy standalone Philips DVP642 seems to work best with those from a compatibility standpoint).

[buzzqw]

my vote for avc (even better ogg+avc in mkv :) )

BHH

[sillKotscha]

no new codecs... but a "advanced - advanced" tab which enables all the nice features of the two already supported ones. Most new mpeg4 standalone player do support qpel and gmc, etc.

that would be very nice... and if you ask me, when it's done you even can enable two dae.dll's as long as they can be removed :)

cheers Sill

[len0x]

Originally posted by buzzqw
my vote for avc (even better ogg+avc in mkv :) )


I'm thinking more of x264+mp4 (with ogg/aac). I doubt that ogm/mkv will ever be supported in AutoGK.

[Sharktooth]

+ ogg?
uhm... mp4 should be avc+aac. that's why it was made.

[len0x]

yeah, not the ogg :)

[Sharktooth]

ogg is good and free but not for mp4. maybe autogk should support matroska and ogg.

[len0x]

I can't support everything at the same time - there has to be some priorities.

[Sharktooth]

well... then AVC+AAC in mp4 would be the right choice.
btw x264 is rapidly changing. there is a vfw and a command line encoder too.

[len0x]

Btw, anyone wants to explorer x246 area? In order to support it - compressibility test has to be nailed down (I have no idea where to start yet).

P.S. I'm following x264 very closely.

[Sharktooth]

then there's a new revision (103)...:)
i was just talking to celtic druid about the fact i should check the status of x264...

[len0x]

I know :)

[len0x]

I'm actually thinking about putting x264 support into GK first (without MP4 container). That way we at least have a test platform.

[Sharktooth]

the latest builds (just updated) are here: http://ebola.gamersrevolt.it/celticdruid/x264/

[len0x]

I have it bookmarked :)

[buzzqw]

i have choose ogg simple because i prefer (or sound) it over aac. Simply.

But AAC+AVC is very, very, very good !

BHH

[len0x]

ok, it didn't took me that long to add x264 support to GK :)
Watching now my first video done with GK and x264.
Will have something for the public to play with later this week.

P.S. But we really need to think about how to do compressibility test with it.

[akupenguin]

The problem is that main concept that AutoGK is working on - is quality of the encode and comressibility test. I have hard time picturing what would be the solution for x264: what can be considered as 100% quality? Is it quantizers 17-18? Or may be still have 0 as 100% but then we should have a different measure of what looks OK: in ASP its 70% (or average quant < 3). What about AVC? I assume you've done a lot of tests and have at least some pointers about how you determine which bitrate to give to a particular source.
If you want something similar to your current method of mpeg4asp (assuming I understand correctly; I don't use GK), I recommend: Encode 1st pass at quant 18. Encode 2nd pass at ~70% of the bitrate, which tends to be around average quant = 20. (A factor of 0.7 would normally cause quant to increase by 3, but the 2nd pass should use better/slower settings than the 1st.)
However, I would not call quant 18 "100% quality" (nor would I measure quality in % at all): There is no maximum quality for H.264. (In high profile, quant 0 is lossless. In main profile (current x264), quant 0 isn't quite lossless, but is only barely smaller than ffhuffyuv.)

[len0x]

Thanks for the info!

Originally posted by akupenguin
However, I would not call quant 18 "100% quality" (nor would I measure quality in % at all): There is no maximum quality for H.264.

Yes, I see that. Its the most problematic bit of x264. If we to compare to to xvid at the the same size - x264 definitely wins by a big margin, but that means that we can go higher with resolution. For DVD sources this would actually mean that we almost always have 720*xxx resolutions if going for more than 1 CD.

What would be the quantizer when you'll be unable to see _major_ difference between encode and actual DVD at this stage? Is it 18 or lower?

[akupenguin]

Originally posted by len0x
Its the most problematic bit of x264. I would call it a problematic bit of mpeg4: I often wanted slightly better quality than quant 2, but the huge gap between q2 and q1 made ratecontrol suboptimal.

What would be the quantizer when you'll be unable to see _major_ difference between encode and actual DVD at this stage? Is it 18 or lower? I'm hesitant to answer that, as it varies so much between people. But for my own preference with anime content, q20 is good enough to have no annoying artifacts. (2pass of course, with different scenes varying from maybe q16 to q24.) And (again depending on content) q20 is also sufficient to fit 2hours at 704x384 on 1 CD, as long as the source isn't too noisy.

[Sharktooth]

what lenox need is something like q2 for xvid.
so the best possible quant for everything without wasting space.


btw, custom matrices were made also to fit the gap between q1 and q2...

[len0x]

Originally posted by Sharktooth
what lenox need is something like q2 for xvid.
so the best possible quant for everything without wasting space.



Yes, "good enough" is not really enough for that I think. I wanna see almost no difference to the original source for most sources at a particular quantizer. Its not something that can vary for different ppl as comparing frame by frame will always tells you the difference. And maximum quality starts when you unable to see obvious difference.

*Edit* what can vary is what percent of that maximum bitrate you find acceptable for youself. That's a variable for different sources and different ppl.

[Sharktooth]

So, any news about x.264 compressibility test and x.264 -> AutoGK integration?
BTW, once AGK goes AVC+AAC in MP4 i think i will start to support it...

[len0x]

GK needs to have stable x264 support first, then we'll see.
What do you mean by "you will start to support it"?

P.S. you should follow GK with x264 development thread...

[Sharktooth]

support = donations = $$$

(PS: i know about the problems in latest x.264 builds, i just wondered if there's a solution for the compressibility test with h.264 quantizers...)