I'm trying to translate an OpenGL renderer into DirectX9. It mostly seems to work, but the two don't seem to agree on the settings for alpha blending. In OpenGL, I'm using:
glDepthFunc(GL_LEQUAL);
glBlendFunc(GL_SRC_ALPHA,GL_ONE_MINUS_SRC_ALPHA);
glEnable(GL_BLEND);
and never actually setting the GL_DEST_ALPHA, so it's whatever the default is. This works fine. Translating to DirectX, I get:
device->SetRenderState(D3DRS_SRCBLEND, D3DBLEND_INVSRCALPHA);
which should do about the same thing, but totally doesn't. The closest I can get is:
device->SetRenderState(D3DRS_SRCBLEND, D3DBLEND_SRCALPHA);
device->SetRenderState(D3DRS_DESTBLEND, D3DBLEND_DESTALPHA);
which is almost right, but if the geometry overlaps itself, the alpha in front overrides the alpha in back, and makes the more distant faces invisible. For the record, the other potentially related render states I've got going on are:
device->SetRenderState(D3DRS_LIGHTING, FALSE);
device->SetRenderState(D3DRS_ZENABLE, TRUE);
device->SetRenderState(D3DRS_ALPHABLENDENABLE, TRUE);
device->SetTextureStageState(0, D3DTSS_ALPHAOP, D3DTOP_MODULATE);
At this point, I feel like I'm just changing states at random to see which combination gives the best results, but nothing is working as well as it did in OpenGL. Not sure what I'm missing here...
The alpha blending itself is performed correctly. Otherwise, every particle would look strange. The reason why some parts of some particles are not drawn, is that they are behind the transparent parts of some other particles.
To solve this problem you have two options:
Turn off ZWriteEnable for the particles. With that, every object drawn after the particle will be in front of it. This could lead to problems, if you have objects that should actually behind the particles and are drawn afterwards.
Enable alpha testing for the particles. Alpha testing is a technique to remove transparent pixels (given a certain threshold) from the target. This includes the ZBuffer.
Btw. when rendering transparent objects, it is almost always necessary to sort the objects to solve ZBuffer issues. The above solutions work for some special cases.
Related
I wish to make an opengl universal texture transparent hack for the DxWnd tool (an open-source program hosted o SourceForge). The hack should work for every program using opengl to render RGBA textures. DxWnd cah hook and redirect all calls from libraries, including opengl32.dll.
I've read and tried to implement all suggestions about making a texture transparent, including enabling GL_BLEND, disabling GL_CULL_FACE and setting glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA). In addition, there's a routine that enforces the alpha bits of all texture pixels.
I expected that, once finished, the result should be a semi-transparent scene, but that doesn't happen.
For instance, the following is a 3d scene from gl hexen II:
and this is the final result, with some textures not transparent and most pixel colors lost:
Just to demonstrate that DxWnd is able to manipulate color pixels (so that this should not be the cause of the problem) this is the same scene with a filter that recolors every texture:
What could be the reason of the problem? How should I fix it? Please, be aware that since DxWnd is hooking a generic program, it may easily have to confront with opengl calls that have an opposite purpose!
What you want is not generally possible just from hooking onto some other application.
You may be able to force blending to be on. But correct transparent rendering is a fundamentally different task from rendering an opaque scene. Because alpha-blended transparency is based on doing per-triangle blending operations with the background, it only really works if you render everything in a back-to-front order.
But as far as the program is concerned, it thinks it is doing opaque rendering. So it's going to render in the order it sees fit to use. Which for more modern applications is probably front-to-back, to take advantage of early depth testing.
And that's the exact opposite order you need to make transparency work. And there's no generic way to control the order of rendering just by hooking onto a few OpenGL functions.
Furthermore, applications tend to try to avoid rendering parts of the scene that are obviously not visible. So if the application thinks that a particular room is not visible because the door to that room isn't visible, then the room and its contents won't be rendered. So even if you could get the order of rendering correct, you'd also need to make the program change what it renders in order to correctly see through stuff.
It should also be noted that doing alpha blending requires that the fragments being rendered have a useful alpha value. But most fragment computations for opaque surfaces will have an alpha value of 1.0. And thus: no blending. And, unless you're dealing with fixed-function OpenGL rendering, or you're willing to manually patch shaders to add your own alpha uniform values, there's no way to change this from outside of the application.
Setting the scene:
I am rendering a height map (vast non-transparent surface) with a large amount of billboards on it (typically grass, flowers and so on).
The billboards thus have a mostly transparent color map applied, with only a few pixels colored to produce the grass or leaf shapes and such. Note that the edges of those shapes use a bit of transparency gradient to make them look smoother, but I have also tried with basic, binary color/transparent textures.
Pseudo rendering code goes like so:
map->render();
glEnable(GL_BLEND);
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
wildGrass->render();
glDisable(GL_BLEND);
Where the wildGrass render instruction renders multiple billboards at various locations in a single OGL call.
The issue I am experiencing has to do with transparency and the fact that billboards apparently hide each-other, even on their transparent area. However the height-map solid background is correctly displayed on those transparent parts.
Here's the glitch:
Left is with an explicit fragment shader discard on fully transparent pixels
Right is without the discard, clearly showing the billboard's flat quad
Based on my understanding of OGL blending and some reading, it seems that the solution is to have a controlled order of rendering, starting from the most distant objects to the closest, so that the color buffer is filled properly in the end.
I am desperately hoping that there is another way... The ordering here would typically vary depending on the point of view, which means it has to be applied in-real-time for each frame. Plus the nature of those particular billboards is to be produced in a -very large- number... Performance alert!
Any suggestions or is my approach of blending wrong?
Did not work for me:
#httpdigest's suggestion to disable depth buffer writing:
It worked essentially for billboards with the same texture (and possibly a specific type of texture, like wild grass for instance), because the depth inconsistencies are not visually noticeable - however introducing another texture, say a flower with drastically different colours, will immediately highlights those mistakes.
Solution:
#Rabbid76's suggestion to use not-semi-transparent textures with multi-sampling & anti-aliasing: I believe this is the way to go for best visual effect with reasonably low cost on performance.
Alternative solution:
I found an intermediary solution which is probably the cheapest in performance to the expense of quality. I still use textures with gradient transparent edges, but instead of discarding fully transparent pixels, I introduced a degree of tolerance, for example any pixel with alpha < 0.6 is discarded - the value is found experimentally to find the right balance.
With this approach:
I still perform depth tests, so output is correct
Textures quality is degraded/look less smooth - but reasonably so
The glitches on semi-transparent pixels still appear - but are nearly not noticeable
See capture below
So to conclude:
My solution is a cheap and simple approximation giving less smooth visual result
Best result can be obtained by rendering all the billboards to a multi-sampled texture resolve with anti-aliasing and finally output the result in a full screen quad. There are probably to ways to do this:
Either rendering the map first and use the resulting depth buffer when rendering the billboards
Or render both the map and billboards on the multi-sampled texture
Note that the above approaches are both meant to avoid having to distance-base sort a large number of billboards - but this remains a valid option and I have read about storing billboard locations in a quad tree for quick access.
I have a model with transparent quads for a beard. I can not tell what triangles belong to the beard because their color comes from the texture passed to the fragment shader. I have tried to presort the triangles back to front during export of the model, but this does not help. So I implemented MSAA and Alpha to Coverage, but this did not help either. My last attempt was to draw the model, with a depth mask off and skipping any transparent data, so the color buffer would have non-clear color values to blend with. Then I would draw the model a second time with depth testing on and drawing the alpha pieces.
Nothing I have tried so far has worked. What other techniques can I try to get the beard of the model to properly draw? I am looking for a way to handle this that doesn't use a bunch of extensions. I'd prefer techniques that can be handled with plain old OpenGL 4.
Here is an image of what I am dealing with.
This is what I got after I applied the selected answer.
What you're trying to do there is a still largely unsolved problem: Order independent transparency. MSAA is something entirely different, as is alpha coverage.
So far the best working solution is to separate the model into an opaque and a hairy part. Draw the opaque parts of your scene first, then draw everything (semi-)translucent, ordered far to near in a second pass.
The way your image looks like it seems like the beard is rendered as the first thing, which is quite the opposite of what you actually want.
Simple way:
Enable depth write (depth mask), disable alpha-blending, draw model without the beard.
Disable depth write, enable alpha-blending, draw the beard. Make sure face culling is enabled.
Hard way:
Because order-independent transparency in renderers that use z-buffer is an unsolved problem (as datenwolf said), you could try depth-peeling. I believe the paper is available within OpenGL SDK. Most likely it'll be slower than "simple way", and there'll be a limit on number of maximum overlapping transparent polygons. Also check wikipedia article on order-independent transparency.
I'm working on an OpenGL 3 renderer for a GUI toolkit called Gwen. I nearly have everything working, but I'm having some issues getting everything to blend correctly. I've sorted the triangles by which texture they use and packed them into a VBO, so with the Unit Test, it basically boils down into 3 layers: Filled Rects with no texture, Text, and the windows, buttons, etc that use a skin texture.
The Filled Rects are usually drawn on top of everything else and blended in, but the background behind everything is also a Filled Rect, so I can't count on that. There is a Z-value conflict if you draw them last (ex: the windows have a textured shadow around the edges that turns black because the background fails the depth test) and a blending/z-value conflict if you draw them first (ex: some of the selection highlights get drawn on top of instead of blending like they're supposed to).
I can't count on being able to identify any specific layer except the Filled Rects. The different layers have a mix of z-values, so I can't just draw them in a certain order to make things work. While writing this, I thought of a simple method of drawing the triangles sorted back to front, but it could mean lots of little draw calls, which I'm hoping to avoid. Is there some method that involves some voodoo magic blending that would let me keep my big batches of triangles?
You're drawing a GUI; batching shouldn't be your first priority for the simple fact that a GUI just doesn't do much. A GUI will almost never be your performance bottleneck. This smells of premature optimization; first, get it to work. Then, if it's too slow, make it work faster.
There is no simple mechanism for order-independent transparency. Your best bet is to just render things in the proper Z order.
I am going through the NeHe tutorials for OpenGL... I am at lesson 8 (drawing a cube with blending). http://nehe.gamedev.net/data/lessons/lesson.asp?lesson=08
I wanted to experiment and change half of the faces to be opaque so that there is always a semi-transparent face opposite to an opaque one and be able to rotate the cube...
I changed the code a little bit, the entire source is there : http://pastebin.com/uzfSk2wB
I changed a few things :
enable blending by default and set the blending function to glBlendFunc(GL_SRC_ALPHA,GL_ONE_MINUS_SRC_ALPHA);
changed the order of drawing the faces and colors for each face.
I set the depth testing
I draw all the opaque faces
I disable depth test
I draw all the transparent faces
Now, it's hard to tell exactly what is going wrong, but it definitely does not look right, I cannot recognize what face is opaque compared to the transparent ones, sometimes some faces do not seem to get drawn when they should...etc...
Seems like calculating what face is in front compared to back would not be trivial (although I am sure possible), I hope there is a way of doing that would not need to do that.
Looking either for what is wrong in my code or whether this is not the right way of doing it in the first place.
If you disable depth testing before drawing the transparent faces, then they will be drawn with no regard for their correct z-ordering. It probably looks like the transparent faces are being drawn atop all the other faces. Leave depth testing on.