Zdepthdof 1.6 Maya 8.5 On A Mac

[lorentz]

Hi i'm trying to use the zdepthdof shader for mental ray on mac os x 10.4.9, on maya 8.5.
it's always worked fine before but now i just keep getting this error:

// Warning: (Mayatomr.Custom) : {userfolder}/Library/Preferences/Autodesk/maya/8.5/mentalray/include/zDepthDOF.mi: no shader library specified nor found in search path, expect manual loading //

I've tried everything i can think of but it just renders black and give this error. The maya.rayrc is different to previous versions, which might be the cause, but i dunno.

HELP!

cheers

[joojaa]

do you have the compannion library of the mi file in ../lib folder? becase without that the entire mi declaration is just worth nothing.

Is the library compiled for your operating sytem? (taking the windows or linux shader wornt workon mac and otherwise around)

[lorentz]

yeah, it's the right version, with everything in the right places.
I think it's something to do with 8.5, cause i've never had this problem before, and i've installed this shader quite a few times now, on different machines with different versions of maya.

[joojaa]

Well its possible thet the library file is compiled for so long ago that you nolonger can do it agninst the newest mentalray maya 8.5 since it happens.

But more that that it may be that your env unverifiable ae a bit off.

In maya 8.5 you no longer should edit the env file since its ste to autoload certain directories.

where do you download the so file form

[lorentz]

hmmm, i've installed it at home now (the problem was at work).

I get the same error, but it works. which is a bit odd.

I've already edited the Maya.env file, so i don't know what to put it back too.

What do you mean by where did i download the so file from? i got it from here:
http://homepage.mac.com/bauer/portfolio\_sh.../zDepthDOF.html

cheers

[joojaa]

You do understand that theres actually nothing in this shader you need, for any purpose whatsoever. Even when you think you do you can easily do this all yourself.

for example taking the comments form the page

Difficult to set objects in mid-distance to be in focus

umm yes only if you dont use floating point depth that the buffer allready did for you. Besifdes why the hell ramp inbetween at all.

Settings not independend of camera clipping planes

Um no, actualy the floating point depth does not change form frame to frame due to clipping plane change. Well teh compositioning app might show you the depth normalized but thet doesnit mean the data is. its actually true distance and 10 usnits far away is allways the sabme value no metter what the clipping plane says.

Depth resolution a limiting factor

Umm no, its a floating point value. Offcourse if you act n tin the compa pplication and normalize it to 8 bit then offcourse, but if somebodsys doing teht then the shader dont help the error was done after the render.

dont trust your eyes, work on what the specs says is there.

Depth of Field difficult to animate

see number 2, well no it isnt, in fact ist much faster to do animation in comp than in mr. wich do you like instant feedback or rerender?

Only one single zDepth channel

Well yes that, but you can capture all the other effects here with that one channel, so in sanse you can generate all the other images with that channel.

So see this is shader that builds upon one of those things that could be called common knowlege of something matter that has a widespread misconception regrading z-depth the program produces.

This misconception springs form the fact that the buffers generated are encoded int somethings not usually seen in 2d applications (they are negative and unbounded), so people see the normalized image and thisnk gee it has thees problems which it off course hasnt got because iyts actually smarter then you think it is, your just relying on your eyes not what you have. Also incidenttaly adobe is one of these parties that didnt get it.

Second problem its radial, in distance however the camera of a 3d object works by trasofring teh imgae planary so the distance a 3d camera produces is not curved at all. Even if you wish it to be it isnt.. offfcourse if thets the effect you want its easily done n comp without rerender so you can get all the effects herein form one render. Without ever adjusting any shader values and rerender

also maya odes allow you to see trasparency trough zdepth

So basically this is a fix for a problem that does not exsist.

Oh ps i had a lapsus i menat the rayrc not maya,env you shouldnt need to edit rghe rayrc in maya 8.5 Also the mi's need identifiers. so all the intallation instructions are totaly obsolete too.

[lorentz]

wow, thanks!

so could you explain how to get zdepth from maya into after fx.

Also, how would you install it then, if the instructions are obselete.

[joojaa]

QUOTE(martinstace @ 04/18/07, 12:19 PM) [snapback]264306[/snapback]
after fx.

ahh well tehts just it After effects is not a comp app, which is apparent when you want to work on the raw data. Its just by far the worst application out there when you work with depth channels, because it doesnt support them, the production bundle AE supports zdata but VERY badly. Same applies to floating point data, and especially negative dataranges, wich is shame its very simple to implement.

(this is actually the MAIN reason for the misconception, becauase adobe just cant do this right)

the above does partly apply to After effects because its lacking fundamental calculation/fileformat tools. BUt yes it can do this you first need to work out the zchannels so that AE can read them.
'
This is also why i dont call AE a comp application because it hasnt got the basic functionality i call comp, namely freedon to use data as it is.

Anyway just dump the point z depth into afloatingpoint tif (all you need is to connect sampler info to a multiply divide node thet ahs the second value as -1 and output that to a surface shgader if you need to force out a aaed output, then manipulet teh t into soem suitable range in AE)

QUOTE
Also, how would you install it then, if the instructions are obselete.

just copy the mi filke to the mi folder and so files in the lib folder. Or loead them with the shader manager.

[lorentz]

QUOTE
Anyway just dump the point z depth into afloatingpoint tif (all you need is to connect sampler info to a multiply divide node thet ahs the second value as -1 and output that to a surface shgader if you need to force out a aaed output, then manipulet teh t into soem suitable range in AE)

sorry, could you explain this a little more,
how do i dump the point z date into a floating point tiff?
I've made a sampler node, and connected this to a mutiplyDivide node, with the second input set to -1,-1,-1.
What do i connect from the sampler node into input 1? none of them seem to give the depth.

thanks for the help.

[joojaa]

pointCameraZ

[lorentz]

i've connected that, and it seems to be working in the shader material sample (i.e. the shader is black at the bottom and white at the top) but when i render it it's just all white.

[joojaa]

yes offcourse the values are unbounded so your pixel values are probably like 34 or something wich is 34 brighter than your monitor can see. Now the trick is to bring in this unbounded floating point image int comp and scale it to whatever size you want.

I told you it is something you never considered before. see we are nolonger in the cofortable what you see is what you get range of images here.

(it takes some filddling with the resources to get maya rendering the floating point image without clipping it up.)

[lorentz]

I added in another multiplyDivide node after the other one, and set it to multiply the original number by 0.01 and i have something which give different values depending on the depth, but it's not very subtle. It doesn't give you many different grays, and close objects are basically the same grey.

[joojaa]

Ok ill go over the concept again. I dont really see why its not so hard to comprehend. Forget everything you ever tought of pictures. Compsitioning is understanding data not necceserily nice pictutres. Thay come when you understand the data.

Pay attention, think.

Z channels are the DEPTH, in your case its the depth range from 0 to distance (0 being infinite). Dont diminish this your dont actually care what it looks like your going to use that measurement torecreate the image. So the picture is now just RAW measurement data, thats what you want it to be! same applies to zchanels maya creates too.

Now you want to create a focal blur mask. So what you do is determine the focal distance in real maya units, (measure from the picture it now has this information) and map that to 0, then you decide that the width of the focal area and set that to 1 all in between theese values something between 0 and one.

You are NOT trying to render out the mask, that would be stupid because the comp application is perfectly capable of calculating the mask for all possible focal distance values you could ever imagine to do. FROM ONE RENDER!

Dont try to force maya to make the decciccions ahead of time the comp application in this case AE is there to do this calculation based on the data.

The entire point s to NOT try to force the data into 0-1 range.

The floating point image has ample resolution to work with. you dont need a inde3xed color image.

[lorentz]

so i wont be able to use it in after fx, no?
i.e. render from maya, bring into after fx, use to blur, done.

[joojaa]

No, see your thinking in limitted terms. the entire idea of compositioning is to defeer the calcualtion fo 3d applicattions amongst other things.

So what your thinking right now is...

obejct -> shader that makes nice gradient -> blur

but the

shader that makes nice gradient =calcualte distaance -> grade image -> convert to indexed color.

Now since the comp applications and AE is MEANT for processes like grade image -> convert to indexed color ist stupid to do it like this

obejct->calculate distance -> grade image -> convert to indexed color ->blur

Bold is whats inside maya. its far better to

obejct->calculate distance -> grade image -> convert to indexed color ->blur

Now the floating point format is much much richer than 16 bit images in i terms of precision, so it really does not matter whatever the range is you dont need to normalize it. After effects 7.0 Professional edition and forward are able to use floating point images, so use it thets what its meant for.

further the object->calculate distance -> is stable whereas the grade image -> convert to indexed color ->blur is not so as long as your objects dont change you can change the blur in AE even without rerendering the image jsut becasue your range setting si so bad.