Ways of determining file size.

[McQuaid]

I'm fairly new to xvid and not sure exactly on how to choose the best file size based on the content.

I know a lot of people here rip dvds so its a matter of choosing the amount of discs, but I am encoding tv captures. So once I've choosen an acceptable resolution, I'm not sure of ways of choosing the best file size based on compression. I notice a lot of people mentioning using filter x or method y gives them a better/worse % of compression. But how is this determined? Do I do 1 pass highest quality to see the file size? Or is this a case when I would use statsreader?

[McQuaid]

Ok just to add more in case I'm not making myself clear.

I did a 1 pass test highest quality and the file came to 104 megs.

How much more compression can I expect out of a 2 pass encoding? a 100 meg file? 90meg?

I would like to go to the nearest value that makes x # of eps fit on a disc.

For ex. 7 eps at 100 megs. 10 eps at 70 megs.

Just thinking there must be a better way than just roughly guessing what a good file size would be.

[digitize]

One suggestion is to use something like gordian knot to determine a file sized based on the bitrate you want.

[McQuaid]

Ok, I'll try again.

Roughly what percentage does 2 pass get you over 1 pass highest quality?

[exo-tech]

in 2-pass, you choose the desired file size. although it may not be exact, in my experiences [when doing anime episodes], they generally tend to be 5-10MB LESS than what i input. you input the file size during the 2nd time around [2nd pass].

[Teegedeck]

Originally posted by McQuaid
I'm fairly new to xvid and not sure exactly on how to choose the best file size based on the content.

I know a lot of people here rip dvds so its a matter of choosing the amount of discs, but I am encoding tv captures. So once I've choosen an acceptable resolution, I'm not sure of ways of choosing the best file size based on compression. I notice a lot of people mentioning using filter x or method y gives them a better/worse % of compression. But how is this determined? Do I do 1 pass highest quality to see the file size? Or is this a case when I would use statsreader?

Yes, you do a first pass and load the produced stats file into Gordian Knot. And 75-65% of the first pass' size is considered a reasonable size for the second pass.

I'd suggest capturing in MJPEG or huffyuv and then doing an XviD two-pass with some noise-filtering to get the best quality/filsesize ratio.

On the other hand, if you reach 100 megs at constant quality=100 already, that's pretty small! (Do you use limited quantizers?) You gotta know yourself whether it's worth the trouble. Perhaps you can stay with 1-pass and just test the custom-quant matrix that trbarry posted as a substitute for noise-filtering?

[yaz]

Originally posted by Teegedeck
Yes, you do a first pass and load the produced stats file into Gordian Knot. And 75-65% of the first pass' size is considered a reasonable size for the second pass.

whoops ... are u sure about it? i've made & watched hundreds of encoded movies looking perfect (for me, of course) but being quite far away from this suggestion. a 2h movie with acceptable audio on 1cd gives a rate of 30-40% & they look good. keeping your suggestion would give stamp-like size or a spread of 2-3(!)cd for such movies . i've also met such like that but they weren't noticably better. ok, they were neither worse :-)
afaik, with xvid (but also with divx5) this limit can be widely loosen. with divx3 i'd be more cautious.
am i completely wrong or did i misunderstand something?

I'd suggest capturing in MJPEG or huffyuv and then doing an XviD two-pass with some noise-filtering to get the best quality/filsesize ratio.
[/B]

... or give xvid a try! if your system's strong enough u can do a miracle on capturing yet.

the bests
yaz

[Teegedeck]

Originally posted by yaz
whoops ... are u sure about it? i've made & watched hundreds of encoded movies looking perfect (for me, of course) but being quite far away from this suggestion. a 2h movie with acceptable audio on 1cd gives a rate of 30-40% & they look good.
Yes, yes, yes, :D you're right. I determin a good quality/filesize-ratio for very indiviual movie, myself, depending on visuals, going lower than 65% very often. But GKnot certainly gives a good idea about compressiblity and as captures easily look like crap even at reasonable quants, I gave a conservative suggestion, here.

... or give xvid a try! if your system's strong enough u can do a miracle on capturing yet.
I wouldn't exactly recommend re-encoding something done with a lossy codec, though!

[digitize]

yeah i filter in huffyuv, then compress to an mpeg 4 codec.