Hi pacodoni,

I think the Convolution3D dll is optimized to run on P4 with SSE instruction set. Maybe that's why it's slow on your machine. What's your CPU? I'm currently encoding "We were soldiers" again with this script and the KVCD-LBR at CQ_VBR=22 and subtitles ( just for fun ):

LoadPlugin("C:\encoding\MPEG2DEC.dll")
LoadPlugin("C:\encoding\Convolution3d.dll")
LoadPlugin("C:\encoding\vobsub.dll")
mpeg2source("K:\SOLDIERS\VIDEO_TS\soldiers.d2v")
BilinearResize(336,192,45,0,630,480)
VobSub("K:\SOLDIERS\VIDEO_TS\VTS_04_0")
Convolution3d (1, 8, 8, 8, 8, 3, 0)
#TemporalSmoother(1,2)
AddBorders(8,24,8,24)

Elapsed time 27 minutes. Time to complete is 3 hours, 55 minutes.
The running time for the movie is 136 minutes. I'm encoding on a P4 @1.6Ghz.
Source position is now at 15 minutes, 47 seconds and file size is 71 MB ( video stream )

-kwag

I'm working in a P3 850mhz, 512 Mb.

Part of the slow was cos i was doing video and audio togheter...

Encoding only video, the time came to 6 hour, so, for a P3 i guess it's doing ok...

BTW, i notice you're using vob sub in Avsynth, you're adding subtittles ?
IF, you are, you use there and forget about using it in VDUB, is it right ?

Pacodoni

Yes. Just run VobSub Configure and do your subtitles. Then use the script like the one I used. That easy

-kwag

Kwag,

My dvd is chocking on the new VCD-LBR. I guessing it doesn't like
the 64 low bitrate. Plays fine on my PC and Laptop. By the way the
quality has improved from the old VCD-LBR. Set CQ=22 and kept
everything the same. Tried Convolution3d in my script. There is
only a slight improvement. This plugin is for noisy captures, but
the blocks and artifacts are mostly due to lack of bitrates.
Your changes are coming so fast that I'll have to wait awhile until
it settles down. Keep it going though.

-black prince

hey Kwag,

I'm encoding "we were soldiers" too with the new temp. and convolution3d same settings as you and my encoding time is 3 hrs 37 min. I'm also using DVD2AVI and the encode I did yesterday, the audio seemed a little out of synch Did you notice any synch problems?

thx,
ren

Hi ren,
Yes. I have the same problem with that movie!. The audio is out of sync, but not at the beginning. I noticed it after half hour or so. I extracted the WAV with Mediator. I have to try another movie. Maybe there's a glitch on that DVD that throws off Mediator. I re-encoded the WAV with SCMPX. I'll try muxing that audio with the video, and see if I have the same results.
The problem is in the mpeg file. I see the out of sync when played in WMP, so I know it's not template related. If I get the same result, then I'll extract the file with DVD2AVI, and encode it again with Headac3he.

@black prince,
I did suspect that some DVD players won't like the 64Kbps low bit rate. Try 200 MIN or so, and see. I tried in all my 5 DVD players with the 64, and it worked ok. But I guess some DVD players are picky

-kwag

Hi Ren and Kwag,

I also had trouble extracting audio using mpeg medator from
"We were soldiers". Finally, I used dvd2avi and everything worked
fine. There were a few other movies which mpeg medator would
not sync the extracted audio properly. I did the same dvd2avi for
them and it worked.

@Kwag
Checked vcdhelp.com for my dvd model and discovered it takes
min=250, max=2400, and avg=1200 for XVCD. Going to try them
this evening.

@Ren
I am using "U571" DVD as source instead of SVCD's to create KVCD-LBR
encodes. There's alot of water and action scenes. The movie is 1hr
56min 14sec, its 16:9, and film. There are scenes that are difficult
for KVCD-X3 to handle. To answer your email, Tmpgenc was not set
to deinterlace.

Thanx

-black prince

Thanks black prince,

I just finished extracting the AC3 with DVD2AVI 1.77.3, and I'm encoding it now with Headac3he.
It's funny, because the movie "Kate & Leopold" was only synched correctly with Mediator, and with DVD2AVI it always failed
Well, anyway, we have two choices. If one doesn't work, then hopefully the other one will

I finished encoding "We were soldiers", and the video stream size is 611,711MB with subtitles, which make the mpeg larger . So I'm encoding the audio at 192Kbps, for a target size of 195MB. Still it's perfect for a single CD. I previewed the .m1v and it looks great. Hopefully when I mux the audio/video it will be in sync.
And now I take a break, because I'm going to watch the "De la Hoya vs. Vargas" Blood fight

See'ya later!,
-kwag

hey black prince,

I was asking about how you deinterlaced your svcd because i was wondering about the best way to process the interlaced frames or even if it's necessary at all

hey Kwag,

That should be a good match- it would be reeeeaaaally cool if you did a small capture of it and posted a small sample encoded with the LBR temp

thnx,
ren

Hi ren,
That was a GREAT fight
Sadly, I didn't record it
I should have

BTW, finished muxing "We were soldiers". Perfect audio/video sync.
For some reason, Mediator didn't like that movie

-kwag

Hi Kwag,

I've been using your templates for awhile, but this is the first time I've posted on the forum.

I encoded A Beautiful Mind yesterday (PAL Australian release 130 minutes), using your LBR PAL template, and the quality was impressive, given the final file size of 699 MB with 160 Kbit sound. I didn't particular want a small file size version of this movie, and it isn't the most challenging movie to encode, but I happened to have it on the HD at the time.

I'm not sure if you have received any PAL posts for the LBR template, but it is very effective with my tests so far. I also travel a lot, and while my preference is quality over 2 - 3 CDs, the single disk solution is great for when I am away with my laptop.

Will you be releasing a new version of the PAL LBR template? The updated NTSC version sounds even more impressive!

Also, just for the record, I have a Conia M-100 DVD player, which has the same hardware as one of the Apex players (1100W I think), but it has different software installed. It works with all your templates unchanged, except KVCDx3 where I up the resolution to 544x576 (PAL) to make it work. I don't need to resort to MPEG 2, like some Apex owners thank goodness. I must admit that I thought you had lost the plot with KVCDx3, but once I went to 544, I was bowled over by the quality, and the file size, as even at CQ 12, it looks great on most normal (not HiDef) TVs.

Thanks for the excellent work,

Den.

Thanks Den,
Here's the updated KVCD-LBR PAL template:
http://www.kvcd.net/KVCD-LBR-09132002-352x288-(PAL)-PLUS.mcf

I haven't updated the download page. Tomorrow it will be up. But here you have it

-kwag

Hi All,

Still testing. Seems Convolution3D(1,8,8,8,8,3,0) is removing some
detail from faces in the background. Going to try some other settings
suggested on Doom. C3D has very good compressibility though, so
it could become a fixed filter in avs files. Tried Sharpen(.1) because
C3D causing some blurriness. This only increased the filesize a little,
but gave better picture quality.

@Kwag

I feel VCD-LBR is very good for PC and laptop viewing, but
STD TV it still needs some work (my opinion). Even sitting 7-8 feet from
TV the artifacts, blockiness is apparent. Close-ups scenes are great.
I guess Iam spoiled by KVCD-x3.

@Ren

Tmpgenc is too slow and dosen't always do a good job.
Most of my SVCD's were created by DVD2SVCD. D2S will use DVD2AVI's
Force Film for mild cases of deinterlacing and for NTSC or FILM less than
95% it uses the AVISYNTH commands below:

LoadPlugin("E:\DVDBAC~1\2-DVD2~1\MPEG2DEC\MPEG2DEC.DLL")
LoadPlugin("E:\DVDBAC~1\2-DVD2~1\INVERS~1\DONALD~1\DECOMB.DLL")
mpeg2source("D:\TEMP\DVDVideo.D2V")
Telecide()
Decimate(5)
SimpleResize(704,480)
TemporalSmoother(2,1)
AddBorders(0,0,0,0)

I hope this will help you. When I get a movie that is difficult to figure
out how to de-interlace, I'll run D2S for a sample clip and look at it's
avs to determin de-interlacing. About 95% of the time it's correct according
to TheWef.

Hi black prince,

I did several tests last night, and came to the conclusion that the Convolution3D makes the blocks show more that if it's not used. Maybe by changing the parameters I was using, which were a suggestion from the readme file, it will work better. The picture looks sharper if the filter is not used. As for the blocks, I am using a CQ_VBR value of 22, and I think I'll fix that into the LBR on the next update. It make a big difference from 20 to 22, and the file size is not that much larger.
Another thing. I have done some tests now with the CQ_VBR set at 22, and I can't see a difference to a standard VCD. Try encoding your same blocky scene with the standard TMPEG VCD template, and the same scene with the LBR. Watch it on your TV. You'll notice that if you have blocks on some parts with the LBR, the same blocks are on the standard VCD. That's how close ( if not the same ) the LBR is to a VCD. I tried that on my HDTV, and was able to verify that there's barely any difference between the two. So if you can see blocks on a standard VCD encode, you will see the same blocks on the LBR. Try it without the C3D filter.
Give that a try.

-kwag

Hi Kwag,

Tried encoding without C3D and vbr_cq = 22, min = 300 (my DVD Player
needs at least 250) and max=1150. Of course the filesize increased by
300,000 and the quality did increase some. Here's the weird thing though,
the more I increased the max bitrate the smaller my final filesize kept
getting. I started from 1150 and increased it to 2500 by 100 each time.
The quality stayed the same but the file kept getting smaller.
For a one minute encode, it went from 4,916Kb to 3,107KB and my
min=300 time. I know this is no longer a STD VCD with these
bitrates and would not be DVD compliant, but my goal is filesize vs
quality. I am ok with this encode for my PC and Laptop, but it seems
a more efficient method is needed to use bitrates during the encoding.
What does bitrate viewer tell you. By the way, my commercial copy
of "Green Mile" was propably mastered from a studio version of the
movie with higher resolutions than a DVD we can buy. It could be as
high as 1024 x 720. That's why I believe is better throughout the entire
movie (GIGO theory).

-black prince

Hey guys

I have back on testing today, and tried the samples with and without Convolution3D, and yeah i could see what black prince ment about the blur detail from faces in the background.
But hasn't see much blockness difference between them.

Black prince did sharpen included on convolution 3D scrypt make a better image compared with the no convolution 3D scrypt ?
What did you think about the two scrypts, what gave you better quality ?

Also, i couldn't have the results that you had when testing the max bitrates. For me the value was always the same,

Kwag Finished Dracula, about 130 min, CQ 25, without C3D and sound 224. The final filesize was 798Mb, the image is cool, and yeah, i could see some artifacts, but almost nothing..

One thing that i notice about C3D, wuas that, in the start or end of the movie, when screen is black and show something as the name of the artists, for example, when encode with C3D, when the name goes off, no artifacts can be seen, but, without C3D, you see those.

C ya.

Pacodoni

Hi pacodoni,
Could you double check and see if the result I'm getting is the same?
Try some part of your movie, where most notable artifacts are seen, but encode it as standard VCD with TMPEG's standard 352x240 NTSCFilm template. For what I can see, there's hardly any difference in artifacts from the standard VCD to the KVCD-LBR encoded with CQ_VBR set to 22.
The is a difference, but it is so minimal, that it can be disregarded.
Just want to make sure other people get the same results I'm getting.

Thanks,
-kwag

@all

We need to use the power of this forum to test new templets
by selecting a DVD movie(s) and having those who are willing to
run tests selected by Kwag to give reports of what's happening.
By starting with a baseline (i.e. hardware, software and movie),
all testers report what their getting (same results or not). Then
when adjustments are made, we can quickly test changes.
Up until now, Kwag and Ren are generally doing all the work.
Only releasing it when he believes it's beyond beta stage. Some
of my testing results are different from others because my source
is mostly SVCD's. Why not create another section for Testing to go
along with the others. To become a tester you must agree to use the
same source, software and at least tell what OS, hardware your using.
Kwag or someone will put out the test criteria and those willing to
help will post results that are the same or different. This is only an idea.

@pacodoni

Sharpen(.1) is what I used. It's all a matter of taste. I can tell you
that it will slow down your encode and increase you filesize. The
picture quality of KVCD-LBR is slightly fuzzy to me and Sharpen give
people and objects a little more definition. I stop using C3D for now.
On Doom9 there are testers trying for find settings for normal movies
since it is very good at compressing a movie. The latest I tried is
Convolution3D(1,4,6,3,5,2.8,0). The trick is to get settings that have
little or no effect, but you get compression better than TS(2,2).

@Kwag

The first statement is purely a suggestion. If you have your own
reasons not to have testers that's fine. About KVCD-LBR, I don't
know how you achieved the quality that KVCDx3 gets, but I been
tempted to take that templet and reduce its resolution to 352x240
and see if it would give the same quality. KVCDx3 is the perfect
blend and produces no blockiness and atrifacts in high action scenes,
water, explosives, dark, bright, etc. Is it possible to use the GOP and
Q.Matrix settings from it, but downsize the resolution. Maybe that's
what your already doing? Anyway keep on doing what your doing.

-black prince

Hi black prince,
Statement #1 is fine, but when the templates are released, they are released as a point of reference to satisfy most people, and from that point, each person can tweak and make changes to their personal taste.
As they are right now, the LBR, they satisfy the primary goal which is for viewing on small ( or not so small ) tv sets, for travelers, and special "convenience" for carrying your movies on a single CD-R. The quality is far above what I thought possible, and is very close to a standard VCD. I have tested now about 4 different movies, and after creating the same section as standard VCD, the result of that same section as LBR looks almost the same. Even on HDTV. The main problem is not the LBR template, it's the 352x240 resolution. If the source is pristine, then the output from the LBR is excelent. But creating mpegs from DVD's at 352x240, we magnify the macroblocks, and we can never achieve the quality of a commercial VCD, because their mastering source was made either from BetaMax digital tapes or some other extremely high resolution and quality. With machines such as Ursa Gold Diamond and Rank Cintel. These things cost around $1,000,000. Slightly out of our budget
As for lowering the resolution on the x3 to 352x240, it will look worse than the LBR. The x3 uses a very high MAX bit rate, and the CQ_VBR is not a linear relationship with the LBR, because of the high resolution of 528x480. That is, the higher the resolution, the lower we can drop the CQ_VBR value. Also, because the x3 uses a much longer GOP, we can use a higher bit rate/CQ combination. The same GOP used on the LBR, even though it creates a much smaller file size, looks really bad and unstable video and artifacts appear everywhere.
As for testers, everyone can make tests and suggestions. I'm always listening. As for KVCD.Net sending BETA templates, special news, what's comming, and what's cooking on the oven, that's only for donators. That's their bonus ( plus all the gifts they get by mail from us for donating )

-kwag

Kwag.

When i said about artifacts, i wasn't talking about a comparison between kvcd standard 352x240, and the new LBR KVCD.
I was pointing to one question, there were almost no artifacts AT ALL.

Sometimes my english ain't too good, so, my bad.

I'm realy happy with my final version of Dracula on 1 CD, it's like i told about this particular movie, it has lots of fog, many night scenes, and this is hard when you put them togheter, you can see more blocks on this type of scenes.
But, using the template, and, taking advantage from the small filesize that produces, i could raise the CQ, raise the sound and make an almost 130 min movie in a single VCD with a great quality, were you can't see blocks in these scenes.

You said you can't see almost any difference between both, i think that LBR was an improvement, first in the motion, second in the filesize, what let you free to explore the CQ, sound, to get quality and third in the compability.

Joining all that, WOW !

So, i wasn't critisizing ( how culd i i've learned almost everything on the KVCD forum )the LBR, au contraire, i think, like i said, it's an improvement, and yes, i tested, as you said to me to do and i keep thinking LBR is better...

Like i said, my Bad, if i express myself wrongly.

Thanks.


black prince

Thanks by the info on Sharpen, it really helps on the C3D effect.
The forum also is very interesting, full of great ideas.
I'll test some new things, including the C3D config you mentioned.
Well, i'll back to encode.

Thanks

C ya guys

Pacodoni