Give us a sample of the new Red Planet Encode kwag!!

__________________
j3llyG0053

Not bad! What's that in terms of accuracy percentage?

I can't believe it either -- it looks better than a regular VCD! What template did you use? What CQ_VBR?

I've now finished the test. It seems that the final file size was ~50MB less than predicted, but I think that the length and complexity of the movie played a part there along with the large resolution (528x576). The permanent subs in the black border area didn't help the work either

Quality is simply excellent. No visible blocks even when looking on the computer display. I must say that I'm lost for words.

Next step: burn and watch the movie again!

Here's the script I used:

LoadPlugin("c:\windows\system\mpeg2dec.dll")
LoadPlugin("c:\windows\system\fluxsmooth.dll")
LoadPlugin("c:\windows\system\legalclip.dll")
LoadPlugin("c:\windows\system\dctfilter_yuy2.dll")
LoadPlugin("e:\windows\system32\vobsub.dll")
LoadPlugin("c:\windows\system\blockbuster.dll")
MPEG2Source("e:\temp\leffat\patton\patton.d2v")
Crop(2,82,-2,-72)
LegalClip()
BicubicResize(496,288,0,0.6)
FluxSmooth(4,7)
Blockbuster( method="noise", variance=1, seed=5823 )
AddBorders(16,144,16,144)
VobSub("e:\temp\leffat\patton\vts_01_0")
DCTFilter(1,1,1,1,1,1,0.5,0)
LegalClip()

CQ_VBR was 11,7.

Here are my results from the overnight encode with the 50/48 formula... Predicted size: 803mb, final size: 784mb, difference: 19mb.

That's quite a shortfall, so in a moment I'm going to try reencoding with 100/48, but first I want to run an experiment...

No no , that's the LBR sample, 352x240 of "The Green Mile". The CQ_VBR for that is 15.21. I'm still encoding that in a slow ( Celeron 1Ghz ) machine. It's now 483MB at 70% done and wanted predicted final size is 662MB. So it's right on track for a correct destination size

-kwag

Here ya go : http://www.kvcd.net/rclip2.m1v
Cut out from actual final .m1v which is 715,978KB.
Uploading now. Give it 3 minutes before downloading. File size ~5.9MB.

-kwag

98.524156360989289985042032302055 %

KVCD LBR, CQ_VBR=15.21

And now I have a suggestion, something we should use as a rule of thumb. In theory, the longer the movie, the more snapshots should be taken to maintain the same accuracy. So let's do this: Let's take 30 snapshots per every hour of video. For a two hour movie, that would be 60, 90 for a three hour etc. I was probably lucky on "The Green Mile" with 50/48 and maybe I should have done 90/48 because it's a three hour movie. Just a thought

-kwag

Hi SansGrip,

That's still below 3%, so that's very good

-kwag

What I'm experimenting with is compensating within KVCDP for the system stream, and encoding my sample strips as an elementary video stream. That way I can remove the *.mpg option from the "select your sample strip" file dialog, which will rule out anyone accidentally selecting a system stream with audio, or whatever.

After that I'm going to try your suggestion of 30 samples per hour to see if that gets me closer .

Just two questions:
1. Where can you download the .dll files you guys are useing?
2. I know I read some were in the forum that some poeple can't run the preditor. What if you can't run the file preditor? Do you just use fitcd with the .dll's then just run tmpg with your new template?
Thanks for the help.

Forgot to mention that I encoded all movies as "ES (Video only)" with TMPEG 2.59. All sample calculations were done the same way.

-kwag

Just for your information, I just finished "Minority Report" at 704x480, CQ_VBR=9.5, the actual file size is 1358.183KB; the predicted file size=1421.364KB. So the factor is 0.95555. I paste the script manually in FitCD as follows:

IL = Framecount / 50
SL = round(Framerate) * 2
SelectRangeEvery(IL, SL)

Was it within average??

Just a tad low. But not bad at all But why such a low CQ_VBR for 2 CDs Was this a dirty (DivX) source? It seems to me that for 2 CDs, the CQ_VBR should be way higher

-kwag

No, it is ripped from DVD, the movie is 137mins, for 2 CD-Rs. (max=2500, min=600)

Did you use FitCD for resizing? And did you use 2 blocks overscan? That would give you a higher CQ_VBR to encode.

-kwag

I'd say something's wrong there. I got a 1h59m movie on one disc at 528x480 with a CQ_VBR of 9.25. Are you using the new GOP?

Just to inform of the status of "The Green Mile", which I started last night on my SLOW Celeron 1Ghz machine.

CQ_VBR for that is 15.21. It's now 637,824KB at 95% done and wanted predicted final size is 662MB.


So it looks like it's going to be right on target.

-kwag

hi Kwag,

could you post your script for "the green mile"(one cd)? I'm still tweaking my script for Pearl Harbor before I encode it and making some progress.

LoadPlugin("D:\encoding\MPEG2DEC.dll")
LoadPlugin("C:\encoding\Unfilter.dll")
LoadPlugin("C:\encoding\Blockbuster.dll")
LoadPlugin("C:\encoding\LegalClip.dll")
AviSource("D:\PEARL_HARBOR_DSC1\VIDEO_TS\PEARL_HAR BOR_d2v-vfapi.avi")
LanczosResize(336,168,0,0,720,480)

ConvertToYUY2()
LegalClip()
Unfilter(-50,-50)
Blockbuster(method="noise",detail_min=1,detail_max =100,variance=0.5)
LegalClip()
AddBorders(8,36,8,36)

I now see the beauty of the Blockbuster filter- blocks are almost gone but I still don't really understand it's parameters Another thing- Lanczos resizer is way way sharper than bilinear so it's the only way to go with the LBR template

I have to catch up on more than a month of not using any filters and not bothering to learn Lot's of reading to do...

-ren

Hi ren,

Here's the script I used on "The Green Mile":

Using CQ_VBR=15.21, MIN=300Kbps, MAX=1,150Kbps and audio target of 112Kbps.

-kwag

"The Green Mile" is done.

Roll the drums .|.|.|.|.|.|.
Here is the result:

Predicted file size: 662MB
Final file size ->>> 660,765KB

Too early for a beer, but I'll have a couple tonight

-kwag