Well, I guess I'll just have to improve the MA script it even further ;)
But then again, if FuPP gives better quality and smaller file size, then I just rather burry the MA script :mrgreen:
( It did serve us very well :cry: ) :lol:
R.I.P. MA script :hihi:

-kwag

:mrgreen:

Great Audioslave! Pics do look awesome, as I told in my sample above I treated a capture with HybridFupps() preset "Low", means much spatio-temporal filtering.

I recognised that this script still does take more time then MA but it behaves much more dynamically, means sharpen "static" objects, clean surfaces (really clean and not plain if a good source is used like dvd) and unsharpener moving objects.

So Karl I think we can't do say something about the compression effectively of that script according to the pics Audioslave posted, cause in the whole stream there will be some fast motion scenes which will be treated heavier in their parts within the frame.
I did not made an encoding till now but many users at doom9.org did report a higher compression factor. ... But lets see...


PS: Also look in here...
http://forum.doom9.org/showthread.php?s=&threadid=70916
... where Didée also made a function ... EDIT: Uops, that one got a diff. purpose ;-)

Let the testing begin :team:

I am too stupid to do it on my own :roll: ,i am just a follower and :bowdown: to you smart people

@ audioslave:
great pictures, very impressive, extreme quality!
:o

:!:

@audioslave,

Could you post the script you used :?:

-kwag

Of course! Here we go:
I pasted the LanczosResize line from MovieStacker and simply copied the values to the Crop() and HybridFuPP() lines. :wink:

P.S. Don't forget to check out the Doom9 forum from time to time. FuPP is constantly changing and improving this routine.

@Audioslave
The way of separating the resizer from fitCD in one crop() and Hybridfupp() cropping/resizing is exactly the way I also did :D

@Incredible
Have you tried the IPP routine you posted the link to :?: I can't get it to work properly. I'm getting all kinds of wierd error messages... :roll:

Nope, as I am still fully drunk of mencoder exersizes I hadn't the time till now :lol:

@audioslave,

Quick test. Change the MA script line to read like this:

ScriptClip(" nf = YDifferenceToNext()" +chr(13)+ "unfilter( -(fmin(round(nf)*2, 50)), -(fmin(round(nf)*2, 50)) ).TemporalSoften(fmin( round(5/nf), 6), round(5/nf), round(5/nf) , 1, 1) ")

Make a quick 10 second encode or so, of the same part you posted the screenshots with MA and FuPP. Tell me your result :!:
Make sure you have the line: MergeChroma(blur(1.5)) on the MA script.

-kwag

Kwag to me it seems you did just decreased the unfilter "max barrier" to 50 ?

But that on the other hand would avoid right blurring on very fast moving/Luma changing scenes? I think at low movement you still get the whole image proportionally to the motion blurred?

This week I builded a version of MA including an added linear adaptive Spatial filtering engine beside unfiltering and tempsoftening.

It worked mega well except that Avisynth does end up with a memory based error or so at the middle of the sample frameserving.

@kwag
I tried your suggestion. But the results are still the same:
HybridFuPP still gives better output quality. And with the changes you suggested HybridFuPP is also faster than the modified MA script with about 4 minutes on my 02:14 clip :!:

Ok thanks.
I guess the increase of the temporal values, screw up the encoding time.
MA script, DEAAAAAADDDDD :mrgreen:
The only thing I don't like about the FuPP, is that it's FUBAR :lol:

-kwag

@kwag

Are we getting new optimal script near future ?

Does this script work for avi -> kvcd conversions too or is it only for dvd ?

I think we should still stay very very cool! :wink:
We should do some more comparisons before saying that something which till now is proofed could be dead that fast. :)

@ Noobs

Means: MA was already updated some tima ago. In here we do test a script/function which does perform in a total different way. So be patient, maybe that all could result in less excitement in a couple of hours :lol:

Have I heard "ffvfw" ? ;-)

@Dialhot
Huh... :?: :?: :?: 8O

Yes you're right, and the current MA script has really served us well. Not to mention that it's far much simpler that the FuPP script ;)
But we'll see, maybe we come up with a FUMANCHU script :rotf:

-kwag

Yep you heared out that "many peoples dropped FFvfw"-factor ;-)

I still got no Probs with ffvfw, neither with mencoder where mencoder is in my encodings MUCH more faster with the same results.

@Kwag

what is "FUMANCHU" or "FUBAR" ? Did I miss something?

:)

I want to give this new HybridFuPP script a try, but have 2 questions. Using this new function with TMPGenc:

a) Should I use the "High Quality" setting or the "Motion Estimate Search" seting?

b) Should "Detect Scene Change" be checked or unchecked?

Thanks,
Icarus

Search google for "La venganza de fumanchu" and you'll get your answer :lol:
But I was relating the word combination: (FU)FuPP(MA)Motion Adaptive a(N)d (CHU)lerias :lol:

As for FUBAR, it's a term used in many programming examples and documentation almost everywhere.
It stands for: (F)uc..d (U)p (B)eyond (A)ll (R)ecognition :hihi:
( Relating it to the complexity of the script, of course. Not to the functionality )

-kwag

I have problem using HybridFuPP_071_b.

It complains about "no MotionMask" function. I saw on waterenterprise that it is included in Mpeg2Dec2 and Mpeg2Dec. The first one does not exists for avs2.5.4 (is that normal ?) so I installed Mpeg2Dec.

Now the message is "MotionMask : no argument named thY1 (hybridfupp line 116) " :-(

Can you help me ?

Hi Phil,you need to get the correct version of masktools 149

http://forum.doom9.org/showthread.ph...0&pagenumber=2

Its about a quarter way down

Damn, there is a new version. I didn't see it.

Thanks :-)

Its weird i was getting no motionmask(blah blah)when i was using a different version of masktools but since i got 149 no more problems

I edited my post.

Here is another place if anyone else is having trouble finding it

http://mf.creations.nl/avs/filters/

I tested the script, and it is great. However, there is a flaw :!:
After every scene change, there are trails of the prior scene, on the scene just after the change.
Nobody saw this :?:

-kwag

@kwag

I can't see any of these flaws you're refering to. 8O Could you describe it more, please?
BTW I'm using "Motion search precision: High quality (slow)" in TMPGEnc. If you're not, please try again with that setting...

EDIT: No matter what "Motion search precision" I set I can't seem to get the flaws you mentioned...

It's not an encoder issue. It's a filter issue, and it's clearly visible.
Not to mention some color bleeding artifacts that I also noticed on some scene changes.

Here, look at these screenshots, and look at the artifact from the first shot, clearly visible on the second shot.
The planet's "halo", is clearly seen on the second shot, but not on the third shot.

http://www.digitalfaq.com/archives/i.../2004/03/1.png
http://www.digitalfaq.com/archives/i.../2004/03/2.png
http://www.digitalfaq.com/archives/i.../2004/03/3.png

I see what you mean, kwag. But no matter how hard I look I can't see any trace of this phenomenon on the movies I've encoded with HybridFuPP... :? Could you please post the script you used?

Wierd :!: :? I'm lost... Do you use the latest version of the filters used in the function? I don't know if it makes a difference but what the heck...

As HybridFupp does do the scaling for you, you just should "split" Bicubic resize into its cropping part:

Mpeg2Source("K:\RED_PLANET\VIDEO_TS\red.d2v")
Crop(11, 60, 698, 360)

HybridFuPP(704, 272, Preset="High", Chroma=False )

... or do not set the resolutions in HybridFupp if using the whole bicubic command as Hyb.Fupp it gots an internal incoming stream width/height detection ... As I remeber when I couldn't resist looking at the avsi inside

;-)

@kwag

This is how my script looks:
And I've put the HybridFuPP.avsi into the Avisynth 2.5 plugin folder.

That doesn't work for me, because Crop needs even numbers. So I get an error message when I try to open the .avs script.
Is there really any advantage to do it that way :?:

-kwag

I have absolutely no idea. :roll: You're the script wizard kwag... 8) :wink:

Whats the big deal when cropping using even numbers as it makes sense, ok maybe loosing "one" px as line ;-)

Well if advantage or not, if choosing Bicubic and its 0,0.6 algorythm you're already affecting/sharpening the Image before entering into Hyb.Fupp! So IF using Bicubic then change from 0,0.6 to 0,0

So I also do correct myself and recomend to let Hyb.Fupp do all the resizing including the usage of crop() before.

I think it really doesn't make a difference.I'm not :mrgreen:

-kwag