Avisynth: GOP 12 for anime? why not 30?
 

I only need answer for this. Why setting GOP to 12 (15) for animes, when GOP 24 (and even 30 for PAL) gives same results with smaller file size? It has some influence on compatybility with standalones?

Long GOPs for anime give more artefacts around sharp edges.

-kwag

This is what I'm talking about. I set GOP for 30 and have preaty less atefacts than with 12. Pure luck or ... ?

Depends what you are calling anime :!:
If you are encoding new cartoons, or computer generated, then those are treated as a movie because you don't have the sharp edges that you find in the animes we are talking about. For example the new Disney movies, or movies like "IceAge", should be encoded with the regular long GOPs because they are computer generated. However, things like japanimation and stuff like that, where you can see the strong contrast on the lines, those are candidates for short GOPs.

-kwag

And how about Futurama :) I'm using AVS25x script for DVD, TMPEG, GOP 30 (and will try a higher GOP). Artefacts are almost invisible even in high motion scenes, and comparating to 12GOP, I could say that they don't exist.

Edit: I must go now, but lately I will make few tests and write some more experiences.

I'm not sure how those are produced :!:

-kwag

Futurama is like the simpsons: hard outline filed with uniform color.
I realy mus go now. Later will describe some more results, but I'm writing truth, It's not a joke.

Hydeus,
you can't see that that generally, cause it depends on your film, the lenght, the treatment, and the original state. So choose what comes out as the best in your case. I think your case is a high AVG Bitrate thing.

I for example did a reencoding on some 480x576 captured Disney "Donald Duck's Christmas" movies on one CDR80, ... amount of movietime was about 110min.
I used a GOP of 1-41-1 !! And everything came out chrystal clear and using Avisynth it appears like a DVD source on the TV, even at 480x576!
No pumping, no artefacts :arrow: clear.

I defenitely don't want to mix up Kwags GOP recommendation! but only saying to you that all this depends on many many things, ... so do your experiments on your own Source and workout and decide by yourself what comes out as the best for you ;-)

Aren't you mistaken there the length GOP and the number of P frames ? It seems to me that when you write "1-41-1" you mean 1I 41P and 1B.

Here we are talking about the maximum number of picture in a group (GOP length). That is the "25" in 1-5823-2-1-25 (to take the tmpgenc settings order) and not the "5823".

Do I make any error or isn't that parameter taht we need to change ? (I always changed it :-))

No I'm not mistaken (sorry I had to be more detailed):

I = 1

P = 41

B= 1

Seq = 1

max GOP = Unlimited

= 84!!! Frames per GOP! (Sounds crazy but I wanted to give an example on a Cartoon (new one) where it ended up very good! Thats what I wanted to point out.

(I also gave you in case of a normal Movie a sample.mpg using these settings in The Indiana Jones Thread of AVG about 450kbit , do you remember? But thats offtopic now))

And the 5382 I know thats the special personal telNumber joke of Kwag.

If you give the 1-5823-2-1 the max of 24 Frames (for example the norm KVCD spec.) that easely could also be written 1-7-2-1-unlimited .... this would also give you a framecount of 24 within a GOP at KVCD specs.

So I was not enough detailed but not mistaken ;-)

And yes we where talking about that maxFrames Value. Sorry.

Okay, that is the point you didn't mentioned and I didn't catched :-)

That's right. Didn't notice at the moment that you were using a 84 long GOP ! I thought "41 isn't so much compared to 25" :-)

That sure ! BTW, the advantage of having a crazy P number is that you can shorten or increase the longer of the GOP by changing the last number without impacting the structure of the GOP itself. It just like make cut and paste of a "IPBBP" pattern.

:-)

I hope I did understand you well (A nice sentence of mine as I often use it ;-) )
Well I personally do my Gop Count by setting only the I,P and B Frames following by a maxGOP=Unlimited. So I do not CUT the IBP Pattern in ints following at the end of the counting.

If you want to keep at least the full structure of the pattern you should use a maxFrames value so that the interval of the pattern is completed (or you could say "round").

And by setting a 1-41-1 / unlimited this got an interval as following:

IBPBPBPBPBPBPBPBPBPBPBPBPBPBP... and so on.

You see in my example I only choosed 1 B frame within the interval/Pattern. Less end-filesize and also a goog effective B Frame intervall as it does work bi-directional. Well, ... watch the sample, but there was also another Matrix of mine used - still betaPhase

ok fellowes :) Thanks for charming fight :litesbr: Thats realy explains me a lot. You propably now me enought and see that I'm newbie and i experiments alot :crashed:
As for everybody whow wants to know long gops (~40 as Incedible geschreiben hat ;) ) + ULBR template + ULBR resolution + MPEG1 for cartoons like Futurama, Simpsons, Disneys oldies and so one = very good results. But as always, try youre self, mayby you get better ;)

After many encodes I read recommendations about GOP lenght.
All those encodes were done with a GOP lenght = 24. In spite of it, I'm in PAL area, I have not problem.
Which is the GOP lenght importance?

There is no real diff between 25 and 24.
I always found 25 more convenient for PAL as that make a I frame each second (and so, a possible chapter entry). But that's all :-)

Note: I do not use "Detect Scene Change" else this theory would have been false.