FREE 5-10% CCE Speed Increase

[LB]

Blah, it's late so I'll make it brief. D/L a program called FITCD somewhere on doom. Create ya'self a vfapi avi file of your d2v, open the vfapi in FITCD and set it to 1:1 input and DVD output and all the right settings for NTSC video. Then copy the resize and addboarders lines near the upper portion of the box that spits out the .avs code. What this does is 1. resize add boarders to your video; then 2. resize it to 720x480.

Eh? Probably confused right? Well, most TVs out there clip your video a few pixels (~8) on each side and this essentially means that some of the video is being cut off. So, add yaself some boarders and this will do three things for you:

1. Speed up encode time in CCE since there is less to render. Remember, mpeg2 can handle black perfectly, unlike divx.

2. Better use of your bitrate. As stated in reason.1, you free up some bits to be better used in each frame.

3. You don't lose any of the picture to clipping.

Welp, there you go. Three great reasons to use this program to auto create your boarders for you. Simple as cake, just paste into your avs what this program spits out.

LB

[gizmau]

you didnt read the manual of fitcd.

you can open the d2v directly, there is no need for vfapi, because fitcd creates an avs which uses mpeg2dec/mpegdecoder/whatyoulike-plugin for avisynth. you dont need to copy anythink, fitcd can even save this avs. ;)

and there is nothing new about this: http://www.uni-kassel.de/~eckhardm/

[LB]

No I didn't read the manual. And so now I know that FITCD can open a .d2v. Hurray. Not like it will save me any time really, but oh well. As for the fact that FITCD can create a .avs file. I already know that. But you know the reason people like avisynth? Because they can put whatever they want in it. I don't want all that extra crap that FITCD throws in, nor do I like to have a lot of null lines since they confuse the heck out of me a few weeks later looking back on it. So, I think the good old copy paste works A+ for just getting those two lines out. (I don't think reading the manual would help here).

Finally, as to your last comment about this being nothing new. Sure it is. I don't read german nor do I see these hints I've given on doom9 or vcdhelp, so I'd say this is pretty new and interesting. ;)

[1loser]

I use fitcd all the time but I don't use it for giving me the correct size because it is wrong sometimes. I'm using ver 1.0.4. Maybe there is a newer version.

An example is fitcd says to use:

BicubicResize(480,384,0,0.6,0,0,720,480)
AddBorders(0,48,0,48)

but the correct script is really

BicubicResize(480,360,0,0.6,0,0,720,480)
AddBorders(0,60,0,60)

Correction, I just noticed something

Right above the destination size there is a slider, if I set it to MMX optimized I get the correct size. I don't understand German so the documentation doesn't help me. Does anyone know of any documentation in English?

[Boulder]

The docs are already in English at least if you downloaded the package from ssh's site. The latest version is 1.1.2. I always crop mod-4 and resize 'block optimized' and have had correct calculations.

[1loser]

Originally posted by Boulder
The docs are already in English at least if you downloaded the package from ssh's site. The latest version is 1.1.2. I always crop mod-4 and resize 'block optimized' and have had correct calculations. I found the instructions in English as a text file. I was hoping for an html version with pictures. What is mod-4, I wasn't able to find that in the instructions. When you use 'block optimized' you actually get a very small difference in the actual size and fitcd even tells you the error as for example 2.2% error for block and 6.67% for macro block optimized.

BicubicResize(480,368,0,0.6,0,0,720,480)#'block optimized',2.2% error
AddBorders(0,56,0,56)

BicubicResize(480,384,0,0.6,0,0,720,480)#'macro block optimized'6,67% error
AddBorders(0,48,0,48)

actual size is be

BicubicResize(480,360,0,0.6,0,0,720,480)
AddBorders(0,60,0,60)

I would like to know the difference between "block" and "macro block" optimized and if this really helps in picture quality.

[Boulder]

Mod-4 = divisible (is that the correct word:confused: ) by four, set the round to sliders in cropping accordingly. This is just to make sure that Avisynth 2.5a won't crash. I don't crop usually crop anything but sometimes it's a must.

The difference between blocks and macroblocks is that one macroblock consists of four blocks. A macroblock is 16x16 pixels and one block is 8x8 and when you put four of these together to form a square, that's one macroblock.

I think that the encoders handle one macroblock at a time. That's why it would be better to keep the macroblock completely black so that there would be no sharp edges. Still, block optimization is a lot better compressibility-wise than no optimization at all. You should also use TV overscan blocks if you plan to watch the video on your TV.

Block optimization also comes in handy when you want to adjust the resizing to get the AR error as small as possible. If you select macroblock optimized, the resolution is changed 32 pixels at a time and as you said, it doesn't produce good results. I adjust the resizing with block optimization selected (+2 overscan blocks) and the AR error has mostly been bit less than 2% which you most likely won't be able to notice.

I hope I've not made too many wrong assumptions but this is the way I think it is. The main thing is that more black->more bitrate to the actual film pixels.

[1loser]

Originally posted by Boulder
The main thing is that more black->more bitrate to the actual film pixels.This part I definitely understand. Thanks for explaining things. :)