help with Avisynth Terminology

[bazz]

can someone explain or link me to some guide to help me understand overshoot and undershoot.
Ive searched gooogle , read the Wiki pages and i can not find any answers.

ive been doing alot of small X264 rips from 720p sources and XviD rips from 720p or DVDR

so ive been trying out different scripts i find on google to denoise and sharpen, running off small samples to decide which is better.

Ive had some success with this one i found on a thread in this forum

FFT3dGPU(sigma=1.5, bt=3, bw=32, bh=32, ow=16, oh=16)
MT("""LSFmod(smode=5)""", threads=4, overlap=8)

this is fine but its about time i understood the scripts better and made my own scripts, in this script above ive been using
ive found all the info on sigma , mode etc but i dont understand
overlap ?
and threads, how many threads can be used?

any help , any pointers would be appreciated..thanks

[Zarxrax]

Of course, "overshoot" and "undershoot" are rather ambiguous terms which can depend on whatever you are talking about.
Since you have lsfmod in your post, I guess you are talking about in regards to that.

The original post for LimitedSharpen has a good explanation of overshoot. http://forum.doom9.org/showthread.php?s=&threadid=84196

[bazz]

many thanks Zarxrax for the link mate, much appreciated:thanks:

threads=4, overlap=8

what does threads refer too?, is it kind of workers/ processors?
and overlap, ive looked here on the FFT3dGPU thread and it mentions overlap but doesnt quite explain it
http://forum.doom9.org/showthread.php?t=89941

How on earth do you guys get your head around this stuff, is there a beginning to this ?

i cant find a guide that talks plain english to the beginner

if anyone found any threads that explained the basics please link me
:helpful:


BTW, while im refering to this script, ive heard to sharpen last but i see other guys resizing last in the script , which is better or is there no difference?

Spline36Resize(720,384) # Spline36 (Neutral)
FFT3dGPU(sigma=1.5, bt=3, bw=32, bh=32, ow=16, oh=16)
MT("""LSFmod(smode=5)""", threads=4, overlap=8)
OR
FFT3dGPU(sigma=1.5, bt=3, bw=32, bh=32, ow=16, oh=16)
MT("""LSFmod(smode=5)""", threads=4, overlap=8)
Spline36Resize(720,384) # Spline36 (Neutral)

[Gavino]

what does threads refer too?, is it kind of workers/ processors?
and overlap, ive looked here on the FFT3dGPU thread and it mentions overlap but doesnt quite explain it
threads and overlap are parameters to MT (http://avisynth.org/mediawiki/MT), not the other filters.

[Taurus]

@Gavino:
You took the words right out of my mouth :p.

@bazz:
Spline36Resize(720,384) # Spline36 (Neutral)
FFT3dGPU(sigma=1.5, bt=3, bw=32, bh=32, ow=16, oh=16)
MT("""LSFmod(smode=5)""", threads=4, overlap=8)
OR
FFT3dGPU(sigma=1.5, bt=3, bw=32, bh=32, ow=16, oh=16)
MT("""LSFmod(smode=5)""", threads=4, overlap=8)
Spline36Resize(720,384) # Spline36 (Neutral)

Resize before denoising/sharpening = faster because of less pixels to crunch.
Resize after denoising/sharpening = slower because of all pixels to crunch.
But........
Sometimes it's better to work with the whole picture qualitywise.
It all depends of what you're after.
Do your own little tests on some small samples from your sources, (trim() is your friend),
to get a look and feel what' happening.
Especially when you are downscaling from Bluray sources
and/or you're multithreading your scripts, sometimes shifting resize from start to end
and vice versa does all the speed gain.
And read some Avisynth docus.
The first hundred meters are the most challenging,
beyond this limit the sun shines day and night:D:p
The Avisynth Wiki is a good starting point.
Or learning by doing.
And another hint:
Limitedsharpen (LSFMod)works nice with SetMtMode(2)

[bazz]

thanks Gavino, that was right infront of my eyes and still i looked for filter references
so MT enables multithreading to make the processing faster , hence threads 4 :)

many thanks to you Taurus, now thats what i call plaiin english :)
im gonna sit down tomorrow with no other distrations and read up on the Avisynth Wiki
im think im trying to run before i can crawl.

i tried this but im getting somthing like 4-6 fps , really slow (Phenom II 940BE)
is this not meant to quicken the process

SetMtMode(2)
DirectShowSource("D:\New.mkv", fps=23.976, audio=false, convertfps=true).AssumeFPS(24000,1001)
#deinterlace
#crop
Spline36Resize(720,304) # Spline36 (Neutral)
FFT3dGPU(sigma=1.5, bt=3, bw=32, bh=32, ow=16, oh=16)
MT("""LSFmod(smode=5)""", threads=4, overlap=8)


please anyone:confused:

[Taurus]

@bazz:
Have a look:
http://avisynth.org/mediawiki/MT_support_page
But dont expect everythings there is written in stone.
All depends on the used plugins and their combination to get full speed.
I had an experience the last days where a script started @98% cpu cycles on all 4 cores.
An hour later all were down to 20% and unusable.
Multithreading is still beta and the combination of hardware and software for this is the holy grail of the gurus here :p:D

[Didée]

i tried this but im getting somthing like 4-6 fps , really slow (Phenom II 940BE)
is this not meant to quicken the process
Yes, it's meant to quicken the process. I tried your very script (however with mpeg2source on a PAL SD input video) with x264 --preset fast, and got 81.5 fps. CPU is i7-860.

- Try to avoid DirectShowSource. Use DGDecodeNV or ffms2 instead.

- Regardless of the Source filter, do NOT use SetMTmode(2) prior to the source filter. (Might work in a few special cases, but in general, it just causes trouble or crashes)

- final speed also depends on the encoder settings (a/o application) you're using, about which you told nothing. (e.g. with --preset placebo, your speed would be quite reasonable...)

- I've found that putting ChangeFPS(last,last,true) immediately after the source filter seemingly is good practice. It's not generally needed, but it seems to make strange breakdowns (like reported above by Taurus) much more seldom.

[bazz]

@Didee

Try to avoid DirectShowSource. Use DGDecodeNV or ffms2 instead.

this is just what i need, now and again ive had some crashes (directshowsource.dll not loaded)

what can i load an mkv with instead of DirectShowSource.

i get an error with this, im i using it correctly?

ffms2("D:\New mkv\Ayrton.Senna.mkv", fps=23.976, audio=false, convertfps=true).AssumeFPS(24000,1001)

thanks

[LoRd_MuldeR]

i get an error with this, im i using it correctly?

ffms2("D:\New mkv\Ayrton.Senna.mkv", fps=23.976, audio=false, convertfps=true).AssumeFPS(24000,1001)

thanks

The FFMS2 functions are called "FFVideoSource()" and "FFAudioSource()", so you probably want something like:

vid = FFVideoSource("D:\New mkv\Ayrton.Senna.mkv")
aud = FFAudioSource("D:\New mkv\Ayrton.Senna.mkv")
AudioDub(vid, aud)
AssumeFPS(24000,1001)

If you don't care about the audio track, this can be simplified to:

FFVideoSource("D:\New mkv\Ayrton.Senna.mkv")
AssumeFPS(24000,1001)

Needless to say that you'll need to have an up-to-date 'FFMS2.dll' in your Avisynth plug-in's directory ;)