Log in

View Full Version : What happens during 1st & 2nd pass?


superyupkent1981
22nd June 2003, 10:21
Hi all,

I have been working with the collections of tools within Gordion Knot now for some time already, but there are still some things I would like to understand.

1. What exactly happens during first and second pass, I see that after the first pass there is a rather small file which becomes larger after the second. What happens here?

2. When I started reading the first DivX ripping guide here, Neutral Bibubic was the only filter always mentioned. Now it has been replaced by Lancos. Can someone explain me the differnces between the several choices presented there? Or perhaps give me a link with more reading about that subject.

Thanks in advance,

Alex

killingspree
22nd June 2003, 11:08
hi,
let's see what i can do (:

1) during first pass divx is compressing the video afaik at a fixed quanitziser of two (the same as quality based - quanitziser). sinze this file would become fairly large, it doesn't write it to a normal avi, but stores the information in another form. for the second pass it takes the information on the first pass then tries to use the best quantizer for a certain scene and still not to excede the average bitrate you set.

2) well it's actually a matter of speed versus compressability (which is again influenzed by the sharpness of the picture).
the fastest resizing algorithm is afaik Simple resize, which in my eyes doesn't yield the best result. Bicubic can be tweaked with different settings to be either sharp, soft or, like in your case, neutral. those sharp, soft and neutral are actually just arbitrary definitions for settings the author of gordianknot has chosen for you to simplify the process. Biliniar resize is another algorithm that just uses the vertical lines to resize, therefor iirc softens the image too. Lanczos resize is imho the slowest of these, but yields also the sharpest picture. this way the most information is retained in the picture, but on the other hand compressability is getting worse.
to conclude with a little recommendation: i'd suggest that if you have enough bitrate to spare, let's say with an average dvd movie and 2 mp3 audio tracks it is save to use laczos. it will yield sharper results,l but especially with xvid you might experience a bit of blockiness in high motion scenes! Divx tends to soften the picture anyway so you will get less blocks but can't retain as much quality either! if you are trying to fit a longer movie onto one cd you should go for neutral bicubic or (in extreme) cases even soft bicubic, but don't expect miracles from these either.

i hope i didn't get too technical, if something is still unclear, feel free to ask

steVe

superyupkent1981
22nd June 2003, 11:47
So, Steve, if I understand you correctly, during the first pass GK or divx merely scans the movie in a sort of log file (or other format), to reference to that file during the second pass in which the actual avi is being written?

Thanks for the clear explanation of the resize filters.

Alex

killingspree
22nd June 2003, 11:52
no, actually in the first pass the movie IS compressed, but with the ligthest compressability possible. you have probably already experienced those 'undersized' files when encoding at a rather high bitrate. this happens because the codec is also limited to a max upper bitrate, characterized in the form of quant 2! and in exactly this quant the movie is encoded during the first pass!

steVe