a few questions... please help

[svdf]

hello i have a few questions. maybe some of them have been asked before, but (and yes i've searched) i want to have just an answer to it because all topics say something else, and i'm getting confused of that...

ok let's start:

1 what's faster: reencoding a divx file to add subtitles (so add them in virtual dub and click: save as avi) or adding them and frameserve them to tmpgenc
2 if saving is faster... what kind of compression can i best do? i have a divx file, and want to save it (maybe with a little bit bigger filesize so that i don't lose very much quality...)
3 what is DC component precision? does it have something to do with my dvd player, or could it improve quality/encoding time?
4 what does the vbv buffer size do? this is the only option that has no tooltip...
5 adding a b pictore in the gop structure (so 3 b pictures) i heard that that will increase the quality, is that true? and does it take more encoding time?
6 "no motion search for still picture part by half pixel" does that mean that if a scene has very low motion, it could shock?
7 do you have more tips? (hehheh... )

well i hope someone would be nice enough, and find some time to give me some answers... i have searched all over this forum, and the forum of doom9.org and everybody gives another answer, or says that you have to search...

thanks

[Truman]

Unless you do a VBR 3 or more pass encode, frameserving should be faster. If you save an avi, use a lossless codec.
Concerning your MPEG questions I have some very interesting links. Hope you understand that. I don't.
http://rilanparty.com/vbb/showthread.php?s=bf9c60c37ad541891af2440fc6a5df6a&threadid=1...0&pagenumber=2
http://www.bretl.com/mpeghtml/VBV.HTM
http://www.dvdmadeeasy.com/glossary/index.html
http://bmrc.berkeley.edu/research/mpeg/faq/mpeg2-v38/faq_v38.html
http://www.mpeg.org/MPEG/MSSG/tm5/Ch8/Ch8.html
http://www.mpeg.org/MPEG/MSSG/tm5/Ch7/Ch7.html
http://www.mpeg.org/MPEG/MSSG/tm5/Ch10/Ch10.html

Intra DC precision = 8 should be enough for SVCD. It may save some bits and you wont see the difference between 8 and 10 if you watch it on TV. VBV buffer size depend on the GOP structure and the VBR range. As long as you stay with SVCD standard and use one or two B-frames between I- and P-frames a buffer size of 112 (224 KB) seems to be OK. I can not see a reason to decrease it.
5- I dont think, that changing the GOP structure from IBBPBBPBBPBBPBB to IBBBPBBBPBBBPBBBPBBB give you a better or worse quality automaticly. It highly depends on the source. I played around with the GOP and got some really weird results, for example with very long GOP's like IBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBB. If your source is DivX, that is heavily compressed alreaddy, you may have more success, just check it out.
6-no idea.
7-no, reencoding a DivX is IMHO not a good idea at all. But quality is a subjectiv matter and many people do not care about a "washed-out" picture.

[svdf]

thanks! that helped a lot...
but how do i get that long GOP structure (IBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBBPBB) how many of each i have to have? because i end with BBP and not with PBB...

well i do care about washed out pictures, and i like to have the best quality (spread over 3 cd's) but you said use a lossless codec, but wich codec is lossless? the uncompressed rgb? but what a huge filesize do i get than???

and how can i do a VBR 3 or more pass encode with tmpgenc? because i can only do a 2-pass encode...

[Truman]

If your GOP end with a P-frame you have checked the "close all GOPs" option. This is not recommended at all. If you close all GOPs the encoder does not have a chance to insert an I-frame to a scene change. The image quality will slightly drop. Yes, closing all GOPs helps you to avoid sync problems when editing the MPEG afterwards. But since you can use the source range filter in TMPG you do not need to edit the MPEG.
Lossless codecs are huffyuv or Pic Video or Lead MJPG, have a look at the tool section. With every compression you will loose details, so after all your file would be compressed 3 times. That's why you should use a lossless codec if you need to do one more conversion. You can not "refresh" the DivX, but you wont loose as well. Yes, you would end with a huge filesize, several GB.
With TMPGEnc you can do a VBR 2pass encode only. TMPG is a high quality encoder (but slow), so 2 passes are enough anyway. Cinema Craft let you select more passes, however 2 passes (plus a 3rd one, the creation of a video information file) are enough in most cases. So if you want to use TMPG in 2passVBR mode, the encoding time of both method are almost equal, if you use CBR of course frameserving is faster.