Yeah, its the expo easing..with pingpong
I had a look at changing it but it messed too much with the stage masks to worry about ..the focus has really been on co-ording the stage masks & getting good quality closeup reflections with a lot still happening in the background.

I had written it originally for just <forward> movement as model had initially approached to center stage from the side ..then going to permanent center as close up..but with the taller models you then lose the head with all the kneeling poses.

Yeh, just had a mess and she was poking out the bottom of the mirror. I think I know of another way by using my reflection shader so will give it a go later and see if i can remix it if you don't mind.

Go right ahead squire..love it 👍
If you find a better easing curve ~ may work too. whatever works.
It really pays to comment out the shaders & color the masking quads or you will go nuts finding Arthur from Martha..lol

Yeh, i just commented them out to see what it was doing, must have been a right ***** to get right.

I think Emu did it by using a duplicate model at the start
animated in from the side <forward>pos & opacity which he then timed out
to permanent negative opacity at the scene center point.
The other duplicate model then takes over cycling pingpong from da center point.

Its not rocket science, but its very fiddly to set up. :-). We could do that, however I haven't tired of the novelty of the models nearly spilling off the stage in different ways yet ..hehe.(jk)

Production do a lot of floor work with the model kneeling which screws with fullscreen close up projects and there's no way around that other than keep the <sittingheight> small overall or animate <pos z or scale>and hope for da best. Maybe they need a <kneelingheight> or re-size these clip segments during the production phase or something,

yeh, just tag them as kneeling, for clips that don't go over the front of the table. I guess they have better things to do than pander to the few of us that make scenes though.

@Z22

Sorry no disrespect but I think you mabey wrong when enableing/disableing antialias within the clipsprite, Load wise too is off when disabled from there without changing the framebuffer. Mabey it's my potato not sure, but it's completely off for me doing so. Feel free to enlighten me otherwise please do😄

The display of that framebuffer will be off but the gpu will still calculate it unless you disable the framebuffer itself.

It may have been something else but when I have set up framebuffers before that were somewhat intensive and if i didn't display the buffer my gpu load was the same as if i were displaying it.

Edit:- just tested it with a scene i am working on. It still calculates the framebuffer if it isn't displayed.
As you have done i turned it on/off in the camera node...Was using 30% gpu with it on, 30% with it off. So yeh... turn off the framebuffers you aren't using for lower gpu useage.

http://www.istripper.com/forum/thread/29408/25?post=584691

Alles über iStripper / Share your FullScreen - Member Created Scenes here

Just a simple remix scene using Emu's awesome Fairground scene. There is a optional Antialias function if you so desire. It's off by default. There is random backgounds used so there 2 is to swap them...

Standing Ovation @D7 , Love it lots 😎👍.
Love the fade out to new color repetitions on the clipnamesprites too.
I put a reciprocal link in da Share thread so new zips do not get
spammed out by linkless discussions there. 👌

@Z22

Well I guess I just don't see a spike in GPU on my end. Even if I don't use a framebuffer section at all, for me it's a same normal load....Weird. When it's on I do see a spike and antialias from the sprite. Either way if people see otherwise just delete out the framebuffer section as you say. No biggie😊

Yeh, the AA shader isn't that intensive so you wouldn't see much of a bump. Unlike my shaders which run like a bag of arse...

http://www.istripper.com/forum/thread/29408/25?post=584732

Alles über iStripper / Share your FullScreen - Member Created Scenes here

http://www.istripper.com/forum/thread/27449/69?post=584733 1st July Update of ET - FractalGarden040TB.scn Masking corrections for deep leg overhang reflections etc. Includes FG040TBMod2.scn with fully...

1st July

Dealt with more inefficiencies in FG40 while masking for
deep leg overhang reflections as shown in da above screen caps.
Repositioned the sky (clouds.fsh) in the code sequence as a
smaller sprite mount.
I used a sprite to mount the clouds.fsh because I don't trust
quads to use the precise same placement on the screen as the
colored sprites I usually use to determine masking placements.

I've left the old sequence as // remmed out and incl.
an ET - FractalGarden040TB_old.scnn for peeps to see
the changes made if they are using 'Beyond Compare'.

Also included a mod FG040TBmod2.scn correcting the offset model
starting pos & overcome the 'spilling off the stage' cycle problem
raised earlier ~ tks for noticing dat @Z22.👍
Edit: oohahh Anna gets a tad closer in the mod, but what the hell..
..just reduce her sittingheight if you must.., hehe 😛😈

@ET, here you go. The most recent working "enhancer"
https://drive.google.com/open?id=1i8BA3nxdilU4mGBaILdTjEumJGHVDr4R

This does cause a red chromatic abeeration like error.

These 2 values in the shader control it. i CBA to tune them to fix the error, its most likely the case that i need to change what it does when the if is true.
float resmulU = (2.99);
float resmulD = (3.01);

In the scene resdiv is just a fast way of setting things up for the different res girls. (scaled girl.y/unscaled girl.y) If i can figure it out i will make a shader to autoset that, but it will have to be a seperate shader.

Oki, I downloaded the above. Will have a play 👍

float diff = (1.00);
float diff2 = (1.00);

control the cut off of the effect. 1.0 being minimum or they will cross over and cancel each other out.

float ColorRO = 0.0 + (0.5 - 0.0) * ( (outputR - (0.0) ) / (0.545 - (0.0) ) );
float ColorGO = 0.0 + (0.5 - 0.0) * ( (outputG - (0.0) ) / (0.544 - (0.0) ) );
float ColorBO = 0.0 + (0.5 - 0.0) * ( (outputB - (0.0) ) / (0.546 - (0.0) ) );

controls the output colour balance. mess with the 0.545, 0.544, 0.546 values to adjust each channels brightness without altering black.

With Lorena 720p cards I get up to uniform resdiv of around 3.0
before the white outline gets too distracting.
Its an improvement, however not really enough to qualify as a 1080
equivalent version commercially at this time methinks.
Will tweek some more and see. Tks

Yeh, the outline is a downside of this version. The point was that you can make them look better than they currently do with a realtime shader, totem using a trained neural network in production can do a much better job.

The N_hance one doent get the outline if its the 2 shader version(has a sharpen mask shader too)

https://www.youtube.com/watch?v=8OY8HFGsbKM

Here's the masked version

https://drive.google.com/open?id=1NXPeTiT2SETr35oAnRW_GiG4xatCMbWQ

Doesn't have the outline problem that MDR does but is overly complicated(partly because i cba to cut it down and partly for testing) Has the same source resolution problem though. needs to be set differently for each source girl res.

This has to be 2 shaders too. Or it increases the if's by 27.

The problem i have with all of them is that the girls have compression errors embeded in them which normally are below percepatable threshold unless you take a screenshot and know what you are looking for. But thusfar my attempts at enhancements enhance those glitches too. I have tried various ways or removing those glitches before the enhance but so far i haven't really been able to fix them.

With the masked version the white line is gone,
however I'm not sure the N_Hance7b.fsh was compiling
as I could not see any difference tweeking both uniform variations in da .scn.
Even tried pathing them separately.
Same thang with ''masked", when I substituted AntiAlias.fsh and tweeked these

framebuffer
{
id: N_Hance
size: 3840, 2160
pos: 1920, 1080
sprite {
source: Girl, 0
source: Noise, 1
source: SharpMask, 2
//shader: NH/N_Hance7b.fsh
//uniform: resdiv, float, 3.0
//uniform: resdiv, float, 6.0
shader: AA/antiAlias.fsh
uniform: FXAA_SPAN_MAX, float, 8.0
uniform: FXAA_REDUCE_MUL, float, 0.125 //1.0/8.0
uniform: FXAA_REDUCE_MIN, float, 0.0078125 //1.0/128.0
blend: true
}

oops, forgot to change the .fsh name in the .scn
correct name is

N_HanceM.fsh

the other is correct.

The AA shader only takes in one texture so it would only process the Girl texture.

When testing sequential shaders its better to use their output in the camera node EG:-

sprite
{
pos: -400, 0
source: SharpMask
}

that will show you what that shader is outputting and if it's working. In this case it will be a mostly white outline of the edges such as eyebrows, lips, clothing.

If things don't look like they are working check the filenames because i'm a numpty. ;/

Tks, oK ! wiil do...I ***** crowding things up in one subnode. Explains why It stops dead when i do that ..hehe
Yeah, the video above is great..next video along talks about them having to deploy ' feature relative '
training algorithms using Generative Adversarial Networks because of what seems like
similar artifact contamination issues just dealing with pixel to pixel mapping. Interesting stuff 😱

https://www.youtube.com/user/keeroyz is a good channel if you into neural networks.

Enhance V3(well... more like V30 something...)

https://drive.google.com/open?id=1rUy2iCZYSuQ7uHZp7KqIbVeg-yECBnaJ

Different version of the sharpen mask, this one should be faster. I remembered to include the noise.jpg texture and hopefully i have the .fsh names correct this time. ;/

Currently set for 3k girls.
change resdiv for low res girls. settings are something like... 1.77/3k 2.3/1080p 2.6/720p 3.5/240p. I did figure them out properly but can't find where i wrote it down oops.

If you haven't guessed by now resdiv is dependant on the scaled girl size and framebuffer size vs the unscaled girl.

oh BTW don't use with AA as FXAA blurs the s**t out of the girls and the point of this shader is to fix the blurring caused by the scaling.

It goes a bit weird when the girls have leopardskin on.

I think the resdiv numbers may be half those actually so 3k will be 0.885

I did mess about with and forgot to put them back

float sharpenHR = float ((storefr)*2.5);// Adds a highlight on bright side of edge
float sharpenLR = float((1.0-storefr)*3.5);// Adds lowlight on dark side of edge.

float sharpenHG = float((storefg)*3.5);// Adds a highlight on bright side of edge
float sharpenLG = float((1.0-storefg)*2.5);// Adds lowlight on dark side of edge.

float sharpenHB = float((storefb)*4.5);// Adds a highlight on bright side of edge
float sharpenLB = float((1.0-storefb)*1.5);// Adds lowlight on dark side of edge.

they were
2.5
1.5
2.5
1.5
2.5
1.5

was just messing to see what happened, if you want the girls to look weird just increast the current setting to something like

20.5
30.5 ect..