Nandub codec selection and blocking problems

[AnimeCollector]

Thanks khp. That's what I'm still having trouble with tho. I did read the guides (thank you for posting them), but its sort...confusing...you see, the original file was 175Mb...yet this number (w/o avi overhead) is 18261742....it seems like its missing a digit or something, because if you divide it by 1000 twice...you get 18.26 and that's a very small value in Meg's.

Perhaps I didn't read well enough....but that's my big dilemma now...

unless I can use that bitrate to guess what the final file size will be. It's in kilobits right?

And Enchanter:
/me dangles a nice juicy episode of Fruits Basket

Guess which one? ^____^

Thanks for all your help guys/girls.

[khp]

Originally posted by AnimeCollector
unless I can use that bitrate to guess what the final file size will be. It's in kilobits right?

Yes, with kilo=1000 which is that divx4/5 uses.

I just had a quick look at divx4log myself, and I must admit that it does seem a bit confusing and I don't really see how this tool could be usefull (sorry about recomendig it to you).

[Enchanter]

Finally, this is one thread that has reached up to 4 pages. Congrats to you, AnimeCollector (For being the originator). We don't often see long threads like this.

p.s. /me drools with the sight of FB. So close and yet so far away.

[AnimeCollector]

Enchanter, it was not my intention of making a long thread...while my original question has been answered, I still can't find a proggy to predict file sizes from the log files. Granted it doesn't contain the audio channel, but that's easy to obtain. All faqs and stuff are done mostly for DVD rips, not for re-encodes.

BTW if you have an ftp I can send it to you. Or if you mirc, find me on irc.cobra.net, on the channel #anime-fury. I exist there as I do in this forum: as AnimeCollector.

The only way to find out how that log proggy works is to check that common box, check the audio size, and check final size and see if there is any kind of correlation. Wish me luck; I didn't do it earlier cuz I had stuff to do and less than 300Mb left on my HDD.

Thanks for all your help guys, I'll keep you posted. Maybe this will be useful for you Enchanter in case you fix up files like me and want a way to get predicatable file sizes, or want to check compressability with your filters

[AnimeCollector]

Sound Volume too low

Hmm I guess I can make this my own personal thread without clogging the main forum with threads upon threads upon threads.

Ok, since you guys have been so great thus far, here's another small problem: The sound on one of my eps is too low. I figure I can just change audio to direct processing, set the compression (to whatever I want, I guess) but here's the tough part: what do I set the volume parameter to be? ppl usually use 100%, but what if that isn't enough (ie not loud enought)? I know this is sorta subjective, but is there another way to bump the volume up in a systematic fashion without having to encode over and over again? Would saving the audio by itself save time and grief? BTW i just tried to save wav, it doesn't work...crashes my nandub.

I guess what I"m trying to ask is: what do I set the gain at to avoid clipping, and how does the formula for calculating this work?

[khp]

Let's make this the longest thread ever

There is a guide that covers what you want to do

http://www.divx-digest.com/articles/normalize_avi.html

You might also take a look at the audio utils section

http://www.divx-digest.com/software/index.html#encoding_audio

Personally I would probably decode to wav with avi2wav and use normalize to find the maximum gain, but ofcause I have never tried this.

[Enchanter]

Here's what I do and still am doing:

Extract .wav file using Graphedit, amplify it by 200% (or 6 dB) in Cooledit and send the amplified .wav file into LAME to create the MP3 track. Nandub will handle the MP3 muxing with the video with little fuss. The sound quality is good IMHO.

[AnimeCollector]

Advanced question: encoding from uncompressed file

Oh wise DivX gods (Enchanter, khp) grant me my request, and share with this divx n3wb the secrets of the codec. Hear my rants, oh wise ones, and after musing over it enlighten me with your divx knowledge.

Haha!

Alright, down to business. I want to start subbing and encoding stuff soon, but there is an issue that confuses me. It has been recommended that I take the raw file (divx), save it uncompressed, work with it, then encode it back to divx. The problem is that the uncompressed data could potentially take up 25 Gigs on my HDD. Given I only have a 20Gig device with all my other stuff on it, this is hard to resolve.

Here's my proposal: break up the divx file into smaller chunks, and work with each chunk individually, then merge them all back in the end. The inherent problems I see with this are:
1) VBR - it takes a look at the movie as a whole and assigns data where it is most needed. If a movie requires a lot of data in the beginning, and I divide it up into 4 equal parts, I should technically make the projected file size of the first parts bigger. And the movie will no longer be able to draw data from the low motion parts. Thus, I don't know where to cut the file at, and I don't know the projected file size of each chunk. Hopefully you guys/girls will know what to do here.

I figure that I could look at the log file, decide where each section of high and low motion takes place, divide it according to these regions, and use 1 pass encoding and take a fraction of the total size based on the data I get out of the log files (ie, if a quantizer is set really high, I could take that number and the time length it occurs and assign a specific amount of space for it) 2-pass might even work, who knows?

Khp, I figure I could use the divx log analyzer, but it seems that I assign random compression levels in certain "zones" that I artificially create...so I'm not sure it would be the best thing to use. Maybe I'm wrong. BTW, here's an analysis on the analyzer:

http://www.everwicked.com/content/Advanced_LogFile_Editing/index.php

[khp]

WOW - that sounds complicated

But I don't understand why you have to work with uncompressed video, but considering that I have zero experience with fansubbing, I could be wrong.

As I understand it you should create a subtitles script with SubStationAlpha and then use the subtitler filter for virtualdub to burn it into the avi.

But again I have zero experience with this, so don't trust this to be the one truth.

P.S. don't forget to vote

http://forum.digital-digest.com/showthread.php?s=&threadid=6156

[AnimeCollector]

Neither do I. The program has to decompress as it re-encodes (divx -> uncompressed to apply filters/subs -> divx) but it seems to make a difference.

Voted. It's gonna be a rather spiked bell curve, and we should already know the average age.

The subtitle script for substation also requires me to time it all correctly i'm assuming. So that once I use substation, all the timing has been done (either manually or automatically; i'm assuming manually) then virtual dub takes care of the rest?

[Enchanter]

I believe you need to manually time it, which is very tedious and the reason why there are so few who are willing to be timers among the fansubbers. Once it's done, you can simply send it into vdub.

That's a good way of advertising the thread, khp. I came up with something similar here...

[khp]

Originally posted by Enchanter

That's a good way of advertising the thread, khp. I came up with something similar here...

Yes, I saw you changed your signature, I thougt about that too.
I also thought of something else, we could start blackmailing newbies for votes, kind of like. Yes I know the answer to that question, but I'am not going to tell untill you vote OK maybe that would be taking it too far.

The subtitle script for substation also requires me to time it all correctly i'm assuming. So that once I use substation, all the timing has been done (either manually or automatically; i'm assuming manually) then virtual dub takes care of the rest?

Yes as I understand it, SubStationAlpha will help you make script that contain all the subs along with the exact timeing information, then you setup a encoding in virtualdub using the subtitler filter, which will then apply the subs to movie.

You can get the subtitler filter at www.virtualdub.org. But don't ask me for help with SubStationAlpha I don't have a clue how it works.

[Enchanter]

I haven't a clue on how to properly use SSA either (even if I have experimented a bit, I'm still not good at it). However, I believe an irc op (Akari, I think) of #soldats @ETG occasionally holds timing classes, so you may want to drop in and 'listen.'

[Enchanter]

I sincerely hope this thread will go further into the fifth page and beyond...

[AnimeCollector]

Driving for the 5th page...

If I can't get SSA to work, it's no big deal; I'm sure there are an abundance of subbing programs out there. My main concern is working with the uncompressed file. When you think about it, you're using a partially compressed file to create a new divx file. In order to read the data from the original file you need to decompress it anyway. So the way I see it, it's the as saving the uncompressed file. Yet it works better. Perhaps it has something to do with the ability to read the entire file better, since you don't have to use as many CPU cycles decompressing then recompressing.

So if you read my last post, it seems like I'll need to manually set all of the sizes up. It seems like some intense log file editing...which isn't too bad. I supposed if I could find a formula for calculating the compression levels of the frames I could do it. Little bit of Java and C++ skills to read the log file and create a histogram...not hard either.

Ok, perhaps you can help me with the anti-shit business on nandub. I have a 350Mb episode of Fruits Basket - nice, high quality. Need to squeeze her down to 180Mb tops. Apply my colour and sharpening filters (whew, only 1 hour per pass now), yet when I check the final output, there is a lot of ringing that now appears (you know, the circly artifacts around edges/lines/subtitles) I take it anti-shit takes care of that, since it compares the quality to the original and decides whether to make it a key frame or not. What are some good settings for anti shit? Perhaps if I post some pictures of the output file and the original you'll be able to tell better. The gain in dB is hard to visualize....like, what exactly is an anti shit of 16? How does that translate visually?

Oof! maybe I should cut my posts up to reach page 6 faster