Soft filter?

Those are pretty nice...I esp. like the light beams in the bottom. The only problem I've encountered placing sprites directly in front of the camera is that it can destroy the framrate if you run your monitor refresh at higher refresh rates.

If I have my monitor set at 1600x1200 with a refresh of 65 and run my game at 1024x768 and then put a sprite right in front of the screen, my fps dips from 75 to 30-45. If I run everything the same but the refresh is at 60, it will remain at 60. Likewise if I run the monitor at 1024x768 and use a higher refresh like 85 or up, and run the game at the same resolution, the fps will dip to similar numbers but remain maxed out without the sprite.

This is one reason why the two screens with the glowy fog in my newest screenshot update have been altered to no longer have that look.

You could add a glow filter by copying the current frame onto a sprite which is rendered in front of the camera and which has the bright-flag set. In static picture manipulation you use the same technique and you get great results.
Strength can be adjusted by the blur radius or the sprite's alpha value.

JOEY.

Hiya All,

Going off what Joey had said I thought I'd try the render to texture method abit.
Keep In mind that I have the effect turned up "full blast" and theres still a few kinks in it.

Heres a couple shots in a crummy test level,
when adjusting the alpha abit and set in a better indoor scene it doesnt look too bad.



Next im going to set up materials
for the sprites, this will allow for
alot more control over color and detail..

I beg you to share!!! !!!

Homey, PLEASE tell us exactly how you did your/Joey's blur effect!
It would like REALLY COOL added to in-game slow-motion effects, or when the player goes into some "special play mode" using "some special powerup."

Julian

To get that effect in the 1st Blitz sample is not very difficult at all. The method I know requires you to render two camera views, 1 that is normal and 1 that is very very very blurred. Then make the blurred one 50% transparent (or less) on a layer on top of the regular camera. It gives that sort of dreamy blurry effect, although the shapes are still distinctly made. I don't know if this is possible with 3dgs, and if it is it I have no idea how to do it Try it out with a screenshot in photoshop, it looks really neat. (I'll post something here if I have the time later)

Is it possible to make a single blurred view in 3DGS? If not the method of two cameras is out of the question.

Joey's blur effect... sounds cool... I'm famous *gg*.

The problem with this technique is that you have to blur the sprite's bitmap with gaussian blur or a similar technique. Otherwise you'll see edges and blurry pixels like in the sample above (not that that'd be bad - I find it's great!).

If you invert the bitmap (without blurring it), you can adjust the level's contrast (by changing the alpha value).

If you invert and blur the bitmap, and then subtract it from the view, you get a dark-dream-like effect (as if you were dreaming of a dungeon).

If you blur the bitmap in direction of the camera movement and add it to the view you get a great motion-blur effect.

Many static image filters work with adding/subtracting/mixing (anyhow)manipulated copies to the pictures. I think that there can be taken out a lot for programming shaders.

JOEY.

Are the light beams in the DreamFilter a result of the technique or are those simply a part of the scene.

There is a great executable demo of the Egypt scene posted by Homey above. here is the link egypt.zip there is no source with this and from the looks of things on the site the author is probably not willing to give this information up (but I didn’t ask so who knows) regardless, with the working demo you can see the idea “render to image, blur image, display image” and it runs very very well tested it on my 533mhz Cpu, 320 mb Ram, 64mb NVIDIA Geforce, with no speed loss.

P.S. the demo flickers a bit when you walk around, just ramp up the size, for the effect and it looks really nice, (and still runs well).

anyone thought of using render to texture on a plane(aligned infront of the camera) and then applying the fur shader? this could do great jobs(but maybe a fps-killer??) i dont know;-)

hey i tried with sprites and here are the results:
glow:

w/o glow:
Well its not perfect but it gives the idea of glow

@ello and everyone else trying to use a render_to_view soft effect:

wouldn't there be render to view weaknesses when using render to view for a soft shader?

the view would be rendered onto the sprite, the sprite is transparent and softens it up however it does it, then (on the next frame), the view is rendered onto the sprite, but the view already includes that sprite, so it increases the effect, and this would continue and eventually lead to a fuzzy white view, right?

Julian

not if u use a nontransparent plane model used directly before your eye

Your probably right, (you ARE the FX master) but could you please explain how that works?

Julian

i thought that you use a variable transparent sprite in front of the view. then you render the camera to the bitmap of the sprite while it is invisible. then you blur the picture (or transform it elsehow - described below) and turn it back to it's previous transparent value. by adjusting alpha and the blur radius, you can control strength of the effect.
with a shader (or by the 'bright'-flag) you can either set addive, subtractive or multiply to the sprite and maybe invert it, too. by combining those effects you get great effects.

examples:

• addive + blur => glow
  
• subtractive + invert + blur => negative glow, dreamy, dark effect
  
• addive || subtractive + invert + glow => control over contrast and brighteness
  
• subtractive + glow => ugly effect (looks like a bad painting ;-)
  ...
  

hope you now understood what i meant.

JOEY.

hm... isn't this a recursive explanation?

to do a blur effect you have to render a transparent sprite in front of the camera with a blur effect.

in my mind the only way would be to use different passes with transparency and slightly different positioning to get the blur (just like the furshader works, but another way of positioning).
i dont think that a blur can be done in a single pass using blurformulas as u dont get the values of the neighbour-pixels

but how would you do it in a multipass-technique? that won't be possible, neither.
how do the rendermonkey-examples work, and how many passes do they use?

A small test map would be nice for us graphic noobs! Whats 2 or 3 mb's...

Here's a new Gamasutra article on realtime glow/blur using DirectX 7, 8, and 9:

http://www.gamasutra.com/features/20040526/james_01.shtml

NICE! theres hope...

Based on the Gamasutra article I got blur working in 3DGS. It uses render to texture so it will probably only work in commerical and pro. You need a 320x240 white pcx named 320x240.pcx in the path. You need to be running at 1024x768.

If you want to increase the blur you need to use a smaller image like 160x120 or 80x60. You also need to change the blurcam size_x and size_y to the same size as the image. Additionally you need to change the blurmap.x to fit the blurmap perfectly on the screen. 277 is perfect for 320x240. 69 is perfect for 80x60. Other resolutions (like 800x600) require a different blurmap.x.

As stated in the Gamasutra article there are artifacts but these can be mitigated by increasing the alpha of the blurmap. Making the blurmap less transparent also increases the blur. There will probably be less artifacts the more you blur too (less at 80x60).

---

Explaination of how it works:

You have Camera and Blurcam. Blurcam takes the exact same picture that Camera does except that it does NOT include blurmap in the picture. Camera has a view entity on layer 1 named Blurmap. Blurmap is a sprite (basically a panel) that is 320x240. Blurmap is transparent. Blurcam renders itself to Blurmap (render to texture). Blurmap (a 320x240 texture) is stretched across the 1024x78 view causing it be "blury" except that blurmap is also transparent and therefore you have the regular view and then the blury view superimposed on top.

Frame rate: The only frame rate hit you take is from the extra 320x240 render per view that Blurcam has to render. The blurrier the less of a frame rate hit you take (because the Blurcam render is smaller).

The frame rate hit I noted was LITTLE to NONE.

---

entity blurmap
{
type = <320x240.pcx>;
layer = 2; // display above view entities with layer 1
flags = visible,transparent;// visible on screen from the start
view = CAMERA; // same camera parameters as the default view
albedo = 0; // gouraud shading from sun position
x = 277; // place 277 quants ahead of the view
y = 0; // center horizontally
z = 0; // center vertically
}

view blurcam
{
layer = 0;
pos_x = 0;
pos_y = 0;
size_x = 320;
size_y = 240;
offset_x = 0;
offset_y = 0;
flags = visible;
}

function main()
{
blurmap.alpha=50;
blurcam.bmap=bmap_for_entity(blurmap,0);
cameramove();
}

function cameramove() {
proc_late();
while(1) {
vec_set(blurcam.pos,camera.pos);
vec_set(blurcam.pan,camera.pan);
wait(1);
}
}

---
No blur:



Blur (320x240):



Blur (80x60):

cool how do ya get the stretch part working? and where did 277 calculation come from?

i Have A6 Pro 6.22 i can see the blur map but it stays at 320 x 240 with the code above

On the blurmap entity, X is the distance from the camera. You should be able to lower X (the distance from the camera) to make the entity bigger in the view. Keep lowering X until if completely covers the view. Maybe A6 is different in how far away X is from the camera.

I cant believe commercial users cant use this..WHAT A MAJOR BUMMER (conitec!).

For Commerical users you can modify it to the following and you'll have motion blur but it is slow (about 10 fps) until I can come up with another method. Can commerical users access camera.bmap at all? ->

function main()
{
blurmap.alpha=50;
//blurcam.bmap=bmap_for_entity(blurmap,0);
cameramove();
}

function cameramove() {
proc_late();
while(1) {
vec_set(blurcam.pos,camera.pos);
vec_set(blurcam.pan,camera.pan);
bmap_for_screen(bmap_for_entity(blurmap,0),1,0);
wait(1);
}
}

---

Do shaders have access to the screen texture?

I'm using three layers (a 320x240, a 160x120, and a 80x60) now in my blur effect. It does reduce your frame rate but it is based on the fill rate of your video card and doesn't have much to do with the polygon count that is on the screen at the same time. There shouldn't be much of a hit with one layer. You will get a hit with one layer when you use particle effects at the same time as that will require more fill rate.

whew! yeah the fram rate is low on that..cool motion blur though thanks

you need to move the blurmap closer to the screen, just so it covers the picture.

Julian

Here is a glow code i made for fun. I just modified the blur code found on this topic (pro only but maybe you could do something with multiple views, dunno ):

Code:

---

entity blurmap 
{ 
type = <320X240.tga>; 
layer = 2; // display above view entities with layer 1 
flags = visible,transparent;// visible on screen from the start 
view = CAMERA; // same camera parameters as the default view 
albedo = 0; // gouraud shading from sun position 
x = 278; // place 277 quants ahead of the view 
y = 1; // center horizontally 
z = 1; // center vertically 
} 

entity blurmap1 
{ 
type = <320X240.tga>; 
layer = 2; // display above view entities with layer 1 
flags = visible,transparent;// visible on screen from the start 
view = CAMERA; // same camera parameters as the default view 
albedo = 0; // gouraud shading from sun position 
x = 278; // place 277 quants ahead of the view 
y = 2; // center horizontally 
z = 2; // center vertically 
}

entity blurmap2
{ 
type = <320X240.tga>; 
layer = 2; // display above view entities with layer 1 
flags = visible,transparent;// visible on screen from the start 
view = CAMERA; // same camera parameters as the default view 
albedo = 0; // gouraud shading from sun position 
x = 278; // place 277 quants ahead of the view 
y = 3; // center horizontally 
z = 3; // center vertically 
}

view blurcam 
{ 
layer = 0; 
pos_x = 0; 
pos_y = 0; 
size_x = 320; 
size_y = 240; 
offset_x = 0; 
offset_y = 0; 
flags = visible; 
} 

view blurcam1 
{ 
layer = 0; 
pos_x = 0; 
pos_y = 0; 
size_x = 320; 
size_y = 240; 
offset_x = 0; 
offset_y = 0; 
flags = visible; 
} 

view blurcam2 
{ 
layer = 0; 
pos_x = 0; 
pos_y = 0; 
size_x = 320; 
size_y = 240; 
offset_x = 0; 
offset_y = 0; 
flags = visible; 
}


function cameramove() { 
proc_late(); 
while(1) { 
vec_set(blurcam.pos,camera.pos); 
vec_set(blurcam.pan,camera.pan); 
vec_set(blurcam1.pos,camera.pos); 
vec_set(blurcam2.pan,camera.pan); 
wait(1); 
} 
}

function activate_blur()//put this in your main function
{
	blurmap.alpha = 60;
	blurmap1.alpha = 50; 
	blurmap2.alpha = 40;
	//blurmap.material=mat_glow1;
	blurmap.bright = on;
	blurmap1.bright = on;
	blurmap2.bright = on;
	blurcam.bmap = bmap_for_entity(blurmap,0); 
	blurcam1.bmap = bmap_for_entity(blurmap1,0); 
	blurcam2.bmap = bmap_for_entity(blurmap2,0); 
	cameramove(); 
}

  

---

A couple of screenshots (copy/paste them to your browser):

http://dimenzije.topcities.com/with_glow.jpg

http://dimenzije.topcities.com/no_glow.jpg

looks fantastic. Who wants to sell his pro version very cheap to me?

shaders produce problems on aome nvidia gpus. my radeon works all out, but when i want to show it a friend with a geforce (fx5600 for example) the render to texture shows up, but with no shader on it. using shaders on such things is really cool:)
i found out that simply passing the vertexshader by mov oT0,v7 ... and only doing some manipulations like negating it works on nvidias too, but if you want mor, say you use normals in the vs nvidias go down

Hello All!

I'm trying out the blurmap effect elaborated by Cellulaer.

It works fine with the initial 640x480 pixel screen resolution.
However, when I press the F5 key to increase the screen resolution,
a strange thing happens:

the monitor goes black and sounds like it is switching back and forth
between resolutions. Finally, the computer gives up, interrupts the
program and puts up an error message. What is even more strange is that
the effect works fine if I change the video_mode variable by hand in
the script itself before running the level.

Has anybody else observed the same phenomenon? Might this be some kind
of interaction between the Acknex camera scripts and the blurmap script?
Or is it something which has to do with the bmap_for_entity command itself?

Thanks!

Thomas

is this shader working with DX9? And is the wikipage updated?

Yes blurmap is working fine with the DirectX 9c provided by Conitec.

I wouldn't know about the wiki page though since I took the blurmap from the original forum post by Cellulaer. However, I did noticed that the fur, toon and hatching shaders posted on wiki do not seem to work in A6.3. On the other hand, they work fine in A6.2 with DX9. One of the reasons why I went back to A6.2.

Question: Is it correct to talk about "shaders" in the case of the blurmap effect since it only uses the bmap_for_entity function to map a view onto a sprite. I'm wondering because, from what I read in the manual, I thought shaders were a special type of rendering at the pixel or vertex level.

Thanks,

Thomas, aka Hondarider

The reason it works in 6.2 and not 6.3 is that 6.2 doesnt use dx9 to run-- it still uses dx8. 6.3 is fully dx9 however.

you can apply load the following fx-file into your rendertarget's material:
blur_fx

So does this really mean that DX9 contains DX8 but with the addition of new functions? I was under the impression that DX9 was new from the ground up.

Thanks Ello!

Hum! This is way out of my league. I am guessing that this code will bypass the processes of creation of a view and the bmap_for_entity statement used to map this view on a sprite.

Am I right, or completely off the target and I have to use blur_fx with the original blurmap effect written in C-script language?

of course you need to render the view to a target first and then apply the blur shader to it