I have tried CQ Matic on the same movie now, three times in a row. Each time it comes up with a completely different CQ value, sometimes differences of 2 to 3 points. Any suggestions on prediction methods?

I keep thinking of (I know this will sound ludicrous) a full movie prediction option in CQ Matic? My reasoning is to allow the computer to code the whole movie until it gets it perfect. All the way to a full 800megs. I would imagine the encoding would be absolutely perfect. Just a thought.

__________________
Well if it's not a wolf, then it's a damn big dog.
- Rabbit, from "The Fable"

2 to 3 points, on what range
These 2 to 3 points could be on a flat CQ area, which will basically produce almost the same final file size. Do you mean encode the complete movie at a pre-determined CQ, find out the file size, re-encode once again with another CQ, and keep doing that over and over again until it hits target
How long you think prediction will take

-kwag

I won't even call this "prediction" anyway

You could actually do this:

1) predict the CQ value normally
2) encode the movie
3) if the video file is oversized/undersized (by a user defined percentage), calculate a correlation factor from (final file size)/(predicted final file size) and predict using (factor*desired file size) as the new desired file size
4) encode with the new CQ value

I've done it in the past, used to work quite well.

That will work some times, but not every time.
If the new calculated CQ value falls in a very steep range, the file size will be either way low or way high.

-kwag

True, but it's better than nothing

Yes

That's why I called it an "option".

I guess that with TMPEG's CQ selection, since you can actually get bigger file sizes with smaller CQ numbers, figuring out the correct CQ value is insanely difficult.

If you do a movie at 75CQ, the file size comes to 700megs, and you re-encode it at 74CQ, you can actually get 710megs, It's too screwy.

Man, I had no Idea that file size prediction was this difficult.

__________________
Well if it's not a wolf, then it's a damn big dog.
- Rabbit, from "The Fable"

Now you know what I went through when I did CQMatic
It's a "hair pulling" experience, from the software point of view

-kwag

Would an "option" for larger file sizes (more area of the film) help with prediction?

__________________
Well if it's not a wolf, then it's a damn big dog.
- Rabbit, from "The Fable"

It's already done that way.
The longer the movie, the more samples it takes.

-kwag

The best way still seems to be using CQMatic, then doing final resize with DVDShrink or Rejig. All my encodes from CQMatic that exceed the desired size seem to be in the 95+ percentage in DVDShrink or Rejig and the reendcodes seem pretty good to me. Many CQMatic encodes only require cutting the credits to fit.

The only way I was ever able to reach near total size without shrinking was to process the entire film with CQMatic, then reduce the bitrate with the percentage of overage. This was a pain in the butt and sometimes required 2 or 3 full encodes.

This can't be called "best" way. And be carefull if you decide to do MPEG1 : DVDShrink does not handle it correctly and doesn't even notice you about that !

I haven't given up on this yet
I'm evaluating other prediction alternatives for CQMatic

-kwag

good news for "overburn team" (me include):
"Nero 6.6 supports for the first time a burning of short Lead Outs. Thus one can save up to 80 seconds when burning and has 12 Mb more storage location."
http://www.ahead.de

I announced that 4 weeks ago and guess who answered to me at this time ?
http://www.kvcd.net/forum/viewtopic.php?t=15344

true!!
i remember that someone had post about that feature(in some place).
........ but forgot that was you here Phil!