2nd pass problem and filters

[clodomiro]

1] Hi, I'm using Gknot with 3.11 (although tonight I'm testing 5.02) and I'm trying to make a first pass, see what the stats file tells me, do some corrections (not resolution) and then use the 2nd pass. The problem is that I'm encoding credits at the default low constant bitrate, so when I open the stats for the second pass, Gknot tells me that the stats file is shorter than the actual d2v. So I guess it doesn't matter because what I'm interested in is the movie's compressibility, but I'm worried that the ecf I generate could be wrong. Should I just keep going with the second pass, as only the movie part will be generated?

2] About resize filters. I really can't tell the difference between using 'bicubic sharp' and 'fast and simple', is it because I don't resize much? 704 wide is the minimum I've used. And also, I guess this simple and fast compressess really well, I'm I right.?

[manono]

Hi-

1. I think the ECF file will be generated for only the stats file, which is only for the movie (without the credits included). So I don't really see how the generated ECF could be wrong.

2. The Fast resizer is SimpleResize, which is a bilinear resizer. When encoding off of a good quality anamorphic DVD, you should be able to tell the difference between a bilinear resizer and a bicubic resizer pretty easily, especially at the high resolutions you're using. Find a closeup of a woman's hair. The bicubic resizer should show every strand clearly, whereas a bilinear resizer might blur them a little bit. Or find a closeup of a man's face. A bicubic resizer will show every pore, beard hair, bump and wrinkle, whereas with a bilinear resizer they might "swim" or move around a bit, or blur, and not be as distinct. And yes, the Simple Resize compresses well, but you don't really want to use it. If the compress test results aren't good, lower your resolution rather than go with a bilinear resizer.

[clodomiro]

Thanks a lot, you've been most helpful. I'll do some tests and see what I think.

[clodomiro]

Now I'm confused. When using divx 5.02, I don't know how to load the first pass stats. I guess it's the one called .mv or the .log, but how do I know the movie's Bits/pixel*frame?

[N_F]

You don't have to load any file. It's loaded automatically after the comp. test is finished and you should see your percentage in the window.

[clodomiro]

I'm talking about first pass stats, not comp test. BTW, the 5% means it takes into account the 5% of the movie, or is it random?

[N_F]

I seem to be confusing 5.02/5.03 and all the new versions lately. I can't remember for sure, but I don't think there is any first pass file you can load when using 5.02 (like you can for DivX 3 and XviD).

Yes, the 5% means it will use 5% of the whole movie.

[clodomiro]

Oh, I thought I could get an exact bits/pixel*frame with a first pass just like in 3.11. And I made a typo, I meant to ask if (1) that 5% meant the first 5%, or a 5% of random frames, if the former is the case, how does it account for the great amount of black at the beggining of movies?
Anyway, now I'm confused, with 3.11 I used the first pass to know for sure if the movie could fit into one or two CDs and if I might fit another audio stream (I'm the only one who speaks english around here), but now I have to rely on the comp test? It's not like I've made a lot of movies with Gknot (around 5) but from that short experience I know that a comp test indicating .190 can turn into a first pass of .129 (Road to Perdition). (2)So how do you guys go about these things in 5.xx?
I would stay with 3.11, because in my tests it hasn't been that different from 3.11 (if it looks like crap in 3.11, it looks like a different kind of crap in 5.05) in the dark scenes with subtle gradients in walls and red faces I'm trying to improve, I've come to the conclusion that if I take stills 5.05 is far better as it produces much smaller blocks, but when in motion they kind of move together, thus being much more noticeable, it seems that qpel improves this situation a little bit, but my pIII450 can't play back qpel divx's at 23.97. so it's not an option here. However I trust reviews when they say action scenes are a different matter, so when I get one of those it seems I'd be better off with 5.

[jggimi]

Well, Clodo... I can answer one of your questions. The comp test at 5% will use SelectRangeEvery(280,14), so it's not random, it's 14 frames every 280.

[N_F]

I've always found the comp. test very reliable.

Anyway, if you haven't already you may want to try XviD. With XviD you can find out the first pass size just like you can/could in DivX 3.xx

[clodomiro]

Yeah, I was going to do that, but since I'm just testing if I really like 5.xx over 3.11, xvid will wait a little longer. Thanks for the suggestion, I wasn't sure xvid did that.