Maya depth shader controlled by ramp?
polycount sponsor
Anyone know how to do this, I'd like my luminance depth pass controlled by a ramp so I can control exactly where I want the gradients to appear, I'd like more than one gradient so the near and far are both out of focus.
Replies
all the way at the bottom. might do what you'd like.
Use a "samplerInfo" node's "Point Camera" to evaluate the distance to the rendered pixel. Then it's just a matter of scaling your range of interest to 0-1 to pass on to RGB coloring of a shader. You can create some interactive objects with distance tools all set up to get a more intuitive way of tweaking your range.
I'm not sure if I can share the course material I made but here's a screenshot of my shader is set up in Maya for this kind of shading. There's some interactive controllers in form of locators to set the measuring range.
What do you mean by multiple gradients?
Either way you should be able to set it up similar to the way i described. Should just be a matter of math.
It should just be a matter of tweaking a similar network as the one I showed above.
While we're on the subject does anyone have any advice on how to do a proper DOF-effect on a still image given that you can output the right Z-Depth info?
Only thing I've ever tried is the photoshop Lens Blur with depth info in alpha channel. But I'm thinking that there might be some better software for this?
I wanted to step up to the challenge of making a maya template scene with interactive near/far planes and a falloff value to set up this kind of render. I created a shader for this where you go by a ramp. Was a fun exercise. Every time I do small projects like this it reminds me why I love working with Maya.
Tried the output with Photoshop, works fairly well the Lens Blur filter.
Let me know if you're interested, I can share it.
I don't see why you want it to go from black in the foreground through white in the center and back to black in the distance though... The filter should deal with this. Normally you just render depth information from white to black or black to white. It's possible to get finer precision doing it with two ramps I suppose but if you render to 32bit images then one ramp should give you enough precision and it means you can play with the distance of the focal point in post instead of re-rendering the pass.
Also for most "natural" looking results (if I'm not mistaken) the depth falloff should be exponential into the distance (as it normally is in a game engine's depth buffer). If you think about it in photographic terms you normally only get tight focus with subjects that are close and the further away the subject is the wider the band of focus. If you have tight focus on a distant subject it looks odd (not that you'd never want to use this effect but it's not what your brain expects).
Tight focus on a far away subject makes it look like a scale model. Google for "tilt-shift" photography for examples of this effect.
p.s. kodde: Love your viewcube modification
p.p.s This is the shader I have used to do it in MR before: http://homepage.mac.com/bauer/portfolio_shaders/zDepthDOF/zDepthDOF.html
Never considered the exponential issue you are describing. Isn't this covered by the way a camera works by default? Should I alter my equations in some manner to cover this?
The CompantionCube ViewCube is awesome!
It's just changing a existing image. Not possible in 2011 if I'm not mistaking.
Yea that's a really nice shader/rig you have set up for doing it actually. Always nice to have objects that you can see in the scene to set distances rather than tweaking and re-rendering. I think the only change I would make would be to have just a near and far clip plane and then have the v-ramp fill between them and have it set to white at the bottom and black at the top and interpolation set to "Exponential Down" (assuming it renders the bottom of the ramp as the closest pixels)
(Setting this sort of thing up gets more complex if you want to make it support transparent materials (especially if they can be refractive :P))
Then to render set the mental ray framebuffer to 32bit (e.g. the "Depth (Float) 1x32Bit" preset in 2011) and render to an image format which supports 32bit channels which can also be read by your compositing package.
(p.s. damn.. If only I wasn't using 2011 I may actually have re-enabled the viewcube for once :P)
The math behind the shader calculates the entire range and sets the ramp positions accordingly so there shouldn't(probably somehow) be any room for errors. Maybe you don't get good enough precision?
Never tried rendering to 32-bit. I should get around to trying it.
I never use the viewcube otherwise. Now I just keep it for the visuals, never click it :P I don't like the animated camera movement you get.
The reason I would set it up that way is so that you are rendering purely depth information. This can be interpreted by the compositing package (After effects, Photoshop etc.) so that you can just manipulate the filter settings to give you the focus point and blur depth that you want instead of having to re-render a new gradient every time you want to alter the DOF.
Rendering depth in a single gradient pass like this is especially useful if you need to animate a focus pull in post. Another trick you can use this method for is that you can colour correct the depth pass and screen/overlay/multiply it to use it for atmospheric fading/fog effects.
Before you run the Lens Blur filter on your image make sure you add the depth pass to your beauty render as a new channel. (use "duplicate channel" from your depth image and duplicate it into the image to be blurred). Then run the Lens Blur filter and in the [Depth Map > Source] dropdown box pick the new channel you just added. From there it's just a matter of tweaking settings. "Blur focal distance" refers to what value on the depth map is considered the focus plane. So if you set the Blur focal distance to 128 then mid grey on your depth map will be in focus and the further away from that value the more blurry your image will get.
That's the main trickyness out of the way
ceebee, I'm looking for a way to do this in post it's much faster than setting up a custom depth of field for every shot as well rendering bokeh in mental ray is not that fast for tuning it.
Or maybe you want to learn how to create it yourself?
EDIT:
The above link was broken, so I removed the linking. Check below for newer link.
Cheers Kodde
No worries.
Glad you like it.
Here's the link for the example scene.
To change the near/far limits just move the bigger planes. To change the falloff you have to select the camera shaped controller and change it's Falloff attribute. Sadly I did not find an easy solution for making falloff planes interactive in the sense of just moving them around. They'd have to update on both side regardless of which you chose to move around. The whole setup follows the camera so just navigate around with it as normal.
So far I've only been able to render this with Maya software. For some reason my... *cough* complex shader setup *cough* with Mental Ray makes Maya crash.
Although if all you want is a Z-Depth pass then I don't see why Maya Software wouldn't suffice. Still bugs me though.
The link works fine for me, the one in post #23. The prior one in #19 is still broken. I can remove it.
You are probably right concerning anti-aliasing. I think you can get it quite close, but you will probably always have discrepancies between different renderers.