Ack. At least I only have another 10 or so to go from this spindle :).

Nope, no overburning. I even had about 1.5 minutes to spare at the end.

Hi there,
I have just gone through this whole thread, just want to make sure if I understand it corredtly, here is the formula, right?

Sampler()
##MPEG size = (( Total frames/MovieTimeInMinutes)/24) * (MPEG sample file size * .98 )

The above is applicable when Max GOP=24, right. What if I want to use it to predict KDVD's size. In that case, the Max GOP=18, so do I just replace the "24" with "18" in the formula? like below:

Sampler(length=18 )
##MPEG size = (( Total frames/MovieTimeInMinutes)/18 ) * (MPEG sample file size * .98 )

You got that right :wink:
Good morning all :)

Edit: BTW syk2cy11, you might want to change the encoding mode on the KDVD templates to CQ_VBR, and then encode with file prediction. CQ_VBR should give you better quality that plain CQ.

-kwag

Hi Kwag,
I forgot to mention that, in fact, I have switched the KDVD template from CQ to CQ_VBR but I don't know if it is the best setting. I just modified from the old CQ to CQ_VBR by changing "Rate control mode" and set the max=5000, min=1000 (for Full-D1). I don't want to push you, but I am waiting for your fine tuning!

Hi Kwag and SansGrip,

I decided to encode Minority Report using LBR (352x240) on 1 CD (800MB).
There’s good news and bad:

Movie Length = 8,712 seconds, Resolution (352x240), Resize=336x224

CQ_VBR=22.21 GOP=1-12-2-1-24 Total File size=726,346,795
Predicted File Size = 730,287,000, %Diff = 0.0542% WOW that’s close!!! 8O
Audio (64kb) + Video = 795,832,097. Amazing!!! :mrgreen:

Here’s the bad news:

Picture quality was poor, lots of Gibbs effect around moving people and objects,
artifacts, distant objects were blurry, :( but dark areas were solid in color (thanks to
Blockbuster noise) and the audio was good (sounds like 128kb). :D This film was in Wide
Screen format and that helped to get it to fit on 1 CD. I don’t believe Full
Screen would have fit without a major drop in CQ_VBR. For movies over 90
minutes I would buy them as wide screen to fit for 1 or 2 CD’s. :?




-black prince

Distant objects usually are blurry at that resolution, even with standard VCDs. But it sounds to me like you need to try to raise the CQ_VBR as much as possible. Try smoothing more and adding less noise (even no noise, if it makes a big difference in file size -- IIRC kwag skipped the Blockbuster step when doing The Green Mile). Also make sure you're using bilinear resize.

I always buy widescreen in preparation for when I can finally afford the switch to 16:9 HDTV. Also they tend to be better quality than the fullscreen versions for precisely the reason you just described :).

Hi SansGrip,

SansGrip wrote:
I decided to cut the credits using DVD2AVI with set points. I commented
Blockbuster noise and unchecked anamorphic on FitCD output. The resize
resolution is 496x352 with AddBorders(16,64,16,64). I reduced the
audio from 112kb to 64kb using Headac3he. These changes increased
CQ_VBR to 24.53. The picture quality is very good with alot less Gibbs
effect, artifacts and some improvement of blurry distant objects. Next
time I'm going to create a 3 minute sample movie to check picture
quality before doing a full encode. :)

-black prince

Why did you have this set in the first place 8O. That's good for making a DVD, but not for VCDs. And it makes your final file size MUCH larger. Maybe you can put back Blockbuster :idea:

-kwag

Hi Kwag,

Kwag wrote:
I've been making this mistake. Thinking if the source is anamorphic
then the output should be the same. I already added Blockbuster
noise back because I miss the dark areas with no blocks. It didn't
reduce CQ_VBR by much. I'm still having fits with FitCD and cropping
borders instead of using Tmpgenc's Clip Frame. I though I understood
how to do this, but my results are not what I expect. This time clip frame
top=64, bot=64, left=16 and right=16, but how to get the same with
using crop is puzzeling me. Maybe you or SansGrip could give more
explaination. :)


-black prince

Ok, so you know the size of your borders. Since the source is 720x480, this means that what FitCD calls the "film pixel" (which really means "frame size without borders") should be set to

720 - (16 + 16) = 688

by

480 - (64 + 64) = 352

FitCD will automatically choose what it thinks is the correct "left border" and "top border" for those values, but it might be wrong if the borders aren't perfectly symetrical. If you notice it says, for example, left border is 8 and you know it's 10, then change the 8 to a 10 in the "left border" box. Ditto for the top border.

At this point FitCD knows the size of the frame without borders and how wide each border is. You can then choose your cropping method ("no cropping", "accurate" and "average" are the most common -- I almost always use "average" myself) and set the cropping "round to" sliders both to 4. While you're at it, set the resize "round to" sliders to 16. Also choose the correct destination size, set your TV overscan blocks, and uncheck "optimize for CCE", "Anamorphic" and "Format conversion".

The values shown for cropping and resizing will now be correct for this particular movie, so either copy the script output or insert the values into your own script.

Here's an example from the last movie I did. Source frame: 720x480. Frame without borders: 706x462. Left border: 8. Top border: 8. Average cropping.

I hit the "Source" button and load in the .d2v file. I then set the "film pixel" boxes to 706 and 462, and the "left border" and "top border" both to 8. I make sure the "round to" sliders are all correct, and select the cropping method I want to use -- in this case, "average".

FitCD tells me that I need to crop to 696x460, then resize to 528x352. I put the following lines in my script:

I then copy the AddBorders line verbatim from FitCD's script output:

And that's it: one resized movie with no bits being wasted on not-quite-black borders and with an aspect ratio that matches the original almost exactly.

Hope that helps.

@SansGrip,

SansGrip wrote:Rounding to 16 and unchecking anamorphic to output is where I
got lost. :D This is very clear. Thanks :D

-black prince

Rounding to 16 is not essential but is a good idea, since MPEG macroblocks are 16-pixel squares. By rounding you ensure you're not "splitting" any macroblocks with your border, since this is inefficient and can lead to artifacts.

Let me know how it goes -- you should see much better results now you've unchecked "anamorphic" for the target :).

What a headache, i`ve been trying to follow the developments in this ongoing thread, playing around as new developments came about, tweaking and adjusting.
The movie I`ve been playing with is Beauty and the beast. I live in South Africa, so most off our movies are 16:9 anamorphic. So beauty is full screen. I`ve been trying to put it on one cd with the 704x576 pal plus template with all the new tweeks,gop filter etc. I get to the point where it just doesn`t look good enough, so I reverted back and used x2pal 1cd template Cq28 min300 max2500 and it looks good on my 74cm tv.

What is Gibbs effect?

To use crop, simply do Crop(left, top, -right, -bottom).. so, in the case above, it'd be Crop(16,64,-16,-64)

Hope that helps!

If I don`t enable anamophic in fitcd, my file size with predict is much smaller :D , but then when I play it in my standalone the pcture is about half the size than the original DVD :cry: width wise, is there anyway to overcome this without enabling anamorphic :?: . My problem is that I have a standard 74cm tv, and If I go true 16_9 it`s to small for me width wise compared to the original.

So what i`m saying is that while you guys are able to use 704x576 to put longer movies on 1cd, I cannot because I have to drop my max and cq-vbr so much that picture qaulity drops to much. So I revert back to Standard x2 with the new Gop, but then the filsize predict doesn`t come close, because if you use normal instead off motion estimate, the filesize actually decreases.

Any help would be aapreciated

I think you have the "Blocks TV-overscan" set too high for your TV. You probably have it set to 2 or 3. Try to set it to 1 and leave "Anamorphic" destination unchecked.

-kwag

thanks, i`ll try that tonight. I just finished reading the cq vs cq-vbr topic,and I have had similar findings before thats why I reverted back to standard x2 1cd template. It encodes quiker as well. Will you be incorporating your matrics used in the plus templates or is it the same :?:

The KVCD matrix is already in all the PLUS, SKVCD, and KDVD templates.

-kwag

ok i`ve tried overscan1 left anamorphic off, but sadly it still creates almost half the normal width file :cry: :cry:

I have another resizing method which seems to work a lot better for me, using strictly TMPGEnc to get the proper resolutions -- very simple to do, substantially faster encode times and perfect aspect ratio. When I get home, where I have access to a fast inet connection and a decent computer, I will post it. It might solve your problem vdermerwedz.

thanks for effort :D , i`ll be waiting :roll:

and very simple?.....i'll be waitimg too! :)