KVCD: Help with downloading sample?
 

Hey guys just me again can someone plz tell me where to find a kvcd sample that i can just download and burn with nero. Dont tell me to look in the kvcd.net page on the left side cause i cant find it.

Your sample: Sample_AVI2KVCDmpeg2_Mencoder_352x576_1968kbps.mpg
http://s11.yousendit.com/d.aspx?id=0...8F0A612611EC3D

Prodater do i just burn it with nero as a non complint cd or anything else. :D

SVCD non compliant.

@Pro

I hope you dont mind but I downloaded this sample as well to check it out.

The quality looks excellent especially seeing as it was a scene with heavy rain from Matrix 3.

It was also fun to see you and Phil battle it out lol. :D :lol: :rotf:

Excuse me ???

This sample is awfull. You should look better !

:arrow: Pixelated everywhere ! (scene change at 2' till 3' - flash light at 5' - flash lights at 10'...) But here, I guess than there is no encoder that can do better works due to the high complexity of the scene
(note : was this sample done in 1 or 2 pass ? Because this is REALLY pixelated ! I never saw that anywhere !)

:arrow: Mosquitoes around almost all edges in "statics" scenes (I put taht between quote because even when people don't move, there is still rain so the scene is not such static). 5', 7', 11'...

:arrow: GRAINY AS HELL ! Welcome to memcoder world !

And I think the color are a little washed out but that's perhaps just subjective.

Well I think this Thread doesnt came with a quality contest purpose, but I do agree with Phil that a very lot of quantization artefacts are having their rumble.

The DCT Blocks and rings/mosquitonoise in that encoding is typically for such kind of scenes incl. flickering luma parts like rain, diamond flickerings, sparkles on surfaces, complex watersurfaces etc.

Well I dont agree that grain is added by mencoder but it seems your source needs a bit pre-filtering.

Exactly such a probleme-scene I got yesterday when doing the FreeEnc/Qenc/CCE comparisons on "KillBill Vol.1" at also 2000kbit but 704x576 and that sample did include the introtrailer from MIRAMAX Corp.

The Encoders (2pass) all did mess up at such flickering luma scenes and the big help here was the simple usage of MAscript, as those luma irritations will be seen by YdifferenceToNext immediately which does result in a well proportional unfilter blurring activity. This blur is already enough for the encoder to end up with visibly less! quantisation on those scenes.
And for shure if less quantization is determined by the encoder to those scenes, also the mosquitos will go away.

But watch out as Matrix Episodes do come with many almost black scenes which causes a sgnificant bitrate drop in libavcodec as known.
Im just testing Blockbuster settings which do avoid such bitratedrops and I got a lot of success last night without noise on other parts than almost totally black scenes, these where almost quantized off (purpose of Blockbuster) and arent recognisable later. That gave Libavcodec a big gain in Outputquality, ... only Problem is that MA and Blockbuster (also motion adaptive filtering performing by low luma frame detection (averageLuma()) ... logically are making any encoder less performing.
Next step will be some CCE testing :D

If your Source comes with grain also here MA does on static scenes the known temporal filtering.

I think also that quality is excellent. "Over tastes there is no nothing writed".
If you analize it frame by frame, you see all that you mentioned. But if you let it run, (better in a TV, of course), you don't see any pixelation neither mosquitoes. Excellent quality, even source is a 2 CD80 XviD file.

Edited: I thing that Dialhot and Incredible opinions are right. I am talking about what you see in your TV.

Sorry Prodater, .... but such quantization artefacts on a 352x576 I do recognise on a normal TV ... because the 352x576 resolution will be upsized to 768x576 pixels on a PAL 625 Lines System TV ... and that means the artefacts will be more than double resized in their width.
So best comparisons on PC are if those samples will be scaled to thier final Resolution like 768x576 PAL or 640x480 NTSC 525 Lines System.

You did allocade almost 2000kbit to Letterboxed 352x576 and thats a lot of kbit seen in the KDVD way, such kbits you could also allocade to 704x576 anamorph.

To be fair i thought this sample was pretty good. Like Pro said watching it on a PC monitor and on a TV are two different things.

I know there were artifacts but come on guys this is only a quick sample to help out smarty and it has to be one of the most difficult scene for mencoder or whatever Libavcodec based encoder was used.

I wasn't expect DVD like quality from this sample but I still stand by what I said and think it's pretty good.

It is an extremely difficult scene to reproduce because of the darkness and the heavy rain and thats doesnt help with the pixelation.

Considering the obstacles Mencoder or QuEnc (Or whatever was used) had to overcome this doesnt to too bad in my humble opinion.

Nice sample Pro. :D

EDIT: Alas I to also still see these artifacts on my TV albeit not as bad as on my PC monitor. However like I said it is a very difficult scene to convert and very demanding I should think. :D

You are right, I didn't make this sample with quality comparation purposes. Smarty ask for a sample and I do it quickly with default settings (I use a batch file with default settings). Tweaking some settings for obtain a sample it finished with 2000kb, and with 352*576.
But thank you anyway for your advises, my knowledges are not so deep as yours (Dialhot and other many people here).
BTW I would like to know how do you see artifacts in luma ("flickering luma parts") and how to see it in chroma. (not joke)

The "flickering" luma parts statement was related to the Source Scene content - not artifacts!

In your sample such luma-flashing/flickering is given by the rain. each frame these luma pixels do change and this means a lot of bitrate, as the purpose of mpeg B/P Frame encoding cant handle that as well as regular scenes, so the bitrate has to rise up. But as we do set limits in avg/maxbitrates, the encoder therefore has to quantize more in that part where too less bitrate is allocated :arrow: DCT blocks. and these can occur in 8x8 or even in 16x16 px depending on the Quantization value which is actually given on that frame.

Well related to Chroma artifacts, such ones you can recognise on YV12 encoding (chroma stairs) very fast or Chroma shifts (if the Chroma and Luma channel do not match, i.E. Luma is 1-2px moved to left ... often seen in captures :arrow: especially on animes you can see sometimes those chroma side shadows).
YV12 is a mega low quality Colorspace 4:1:1 is even worse (i.E. NTSC DV)
a quater chroma resolution to full luma resolution. Now you can imagine what happens if you got a 352x288 movie in YV12:

You effectively got
352x288 Luma resolution
176x144 Chroma resolution

at YUY2 you would have
352x288 Luma resolution
176x288 Chroma resolution

Now you do say "but I dont see any chroma stairs" ..
ok, cause a) your Eye wont recognise that chroma loss in comparison to luma loss (the purpose of YUV chromaspace usage) as chroma is the key compount for your eyes. But do watch a YV12 352x288 known JamesBond Intro where the totally red blood goes down the CameralenseHole where our James looks at you with the gun in the hand. Then you will see what I mean.
And b) ... Im not shure but if you do preview those movies they will be maybe converted to other Colorspaces like YUY2 or especially RGB 24/32 and that means chroma interpolation which results in washy colours.
So if you got a 352x288 YV12 Source like a mpeg4 you should keep the colorspace as you can, cause a change to YUY2 like for filtering purposes would mean a quality loss and if then finally a forced colorspace conversion in the encoder will be done (like in TmpgEnc RGB24) then your colourfrequency! (not luma!) would be first resized from 176x144 YV12 to 176x288 YUY2 and then to 352x288 RGB24 ... where the encodec Chromafrequency finally ends up in 176x144 YV12 again.
Now you can imagine that this WILL end up in chroma artefacts like washy colours. And thats also a part of the reason why colorspace conversions do take a recognisable extratime.

Excuse me but I play it normally once in a window taking only one quarter of my 21" monitor and all what I told jump into my eyes !
I'm at office, I can't play it in full screen and I don't have the time to analyse a sample frame by frame (I never do by the way, and I NEVER used any zoom to judge a sample quality, I use only my eyes).

I really think that people have bad idea of what "having a trained eye" means ;-). And trust me, sometimes I would like to have a "normal" eye just enjoy the show :-)

But I agree with your las post : the quality wasn't the goal. I wanted just to clarify the things for Zyphon.

Then excuseme Dialhot, you description was so detailed that I thought that your analisys was frame by frame, and really I wasn`t observed my sample with so critical eye, even I was observed artifacts that you point at, in one of your first post.

hy boys!
now i'm curious to see the quality of that file but i can't download!?!? i only get a file called: "direct.aspx" (4,65kb)
any other link to get it?

That's a yousendit limitation : no more than 20 DL. Or perharps you ussed a DL manager ?

BTW, Pro can you resend it with yousendit ?

thanks Phil, was the dl manager....i got with "normal" IE download.
@ all:
i agree with Phil and ink about the quality(is too bad with mosquitos,blocks, pixelations) and more:.....wrong matizes and a little unsharp....not good for my taste.
sorry,....i have kvcds(mpeg2 480x480) better than this sample(kdvd?)

Actually, no encoder can do better with 1900 KBits as average. This scene is really complex.

But it can be a good idea to use it as a comparison test. Pro if you can encode this same sample with tmgenc , with the MA script, using a CQ value that give almost the same filesize, it would be great.
Use 64 - 2500 as bitrate range.

But if you don't have the time, we won't flame you.

ok! (all)my kvcds mpeg2(480x480) are using min 300 and max 2500(better than 1900)
i have time and want to do some tests.....i need that source(if you elect matrix3 as source)or other...your choice.
send me part of the vob and tell me all you want to test: encoder, script,min&max...everything! i was thinking that we all could use the same source to test all encoders(scripts and filters) a few days like you.....seems that this ideas are "in the air"!

I'm sorry but I only have this sample:
http://www.yousendit.com/d.aspx?id=4...AEC21117FBCEE4
Is a XviD sample, and I have not now the original DVD.
I could make a TMPGEnc sample but not make sense to apply MA script.

I forgot this !

So let's drop the idea : crap in, crap out. That's explain a lot...

yes......like in my signature: "still a law: bad sources takes to bad encodes"
we need to use dvd sources to test encoders and filters....all parameters using good sources!