questions for relighting in post

#1

Hi,
I’ve a question about making use of normal output from maya rendered images in order to relight the 3d rendered objects in Fusion 5.
I’ve seen a plugin for shake which enable the user to map light probes to their image normals.
http://homepage.mac.com/bauer/portfolio_pages/portfolio.6.html

I would like to know if there’s anyway that we can make use of it in fusion 5?

I’m currently making use of channel lighting in Walker Effect plugins to do the trick. But that’s only limited to 3 light source which is RGB only.

#2

I dunno if that is the correct workflow, but that is how i did it…

combinded my image which i want to relight with the normal map through a channel boolean (rgb = xyz)

put the result on a 3d plane, added a light, added a render node… enable lighting in the render node… voilaz… u should see some lighting on the image based on the normals…

hope that helps

#3

would not be necessary the use of depth information (Z-depth)? For calculate the position of lights ?

#4

hmmm… interesting.
Never thought of using a 3d plane for that. but thanks for the idea. i’ll definately try that out.
As for depth map. I’m not sure how useful it is but i’m usually just use it as fog layer or dof in post.

#5

That relighting stuff in post is sooo interesting! :bounce: I know Pixar also used it allot for there ‘Cars’! Some time ago I found a demo-video of it! :slight_smile: Really cool to see! If you are interested, I could search for the URL…!

A guy on the forums here it developing a similar thing for After Effects! Really cool!

combinded my image which i want to relight with the normal map through a channel boolean (rgb = xyz)
put the result on a 3d plane, added a light, added a render node… enable lighting in the render node… voilaz… u should see some lighting on the image based on the normals…

How does the normal map then calculate? Do you mean that the lighting only reaches the parts of the normal map you want? Cause lighting plane with an image on it, can never result in the same lighting as a 3D object… Isn’t it?! (in that case you should Z-displace the plane with a ZDepth map?!)

I can show you something interesting that CMI-VFX recorded on a Flame-box!: CLICK (.mov direct link!). It shows really cool what a ZDept map can be interesting for.
The stupid thing is that only Flame does have that feature… DF of AE don’t have a standard ability for ZDept-deforming. Unless that there are plugins for it (CLICK for one of them :wink: ), its not very simple…

So thats why Normal-maps will do the job better…

I think that playing with the curves will work well in DF. Using óne ColorCurves for filtering the direction of light out of the normal map, and then using that as a mask for a ColorCorrector (so that you can ColorCorrect only the part thats lighted from ‘that one direction’).
I haven’t yet tried that trick in DF, but I’m sure it’ll work well!
[edit:] On that website you mentioned yourself I found a veeeery interesting movie about normal mapping and the abiblities… (you might have seen it yet :smiley: ) The only part I don’t get is that mapping of a UV-layout on the image, ín post… :shrug: If you understand that part, please tell me how it works…! Would it be a sort of distortion thing that refers to a color-map? :smiley: And… even without that plugin you can get very nice results, isn’t it?!

Good luck :wink:
-Gijs

#6

With the new 3D stuff in Fusion 5, you can just export your 3D model into fusion and then re-light it, along with the ability to do shadowing (which a normal/position pass won’t give you)

#7

Hmm… Not everyone is rich enough to update to F5… :wink: And… can you get the same results by using your method? Cause if you have a VFX-scene with a 3D character, I’m sure you can’t get that animated scene into Fusion…? So then you have that same problem again!?! :scream:

About that shadows… Yeah, thats always a hard point, do I think… Is there really no other way then getting the 3D scene, to get you shadows right? (not ‘a sort of normal map’, but then specially for shadowing?)

And Hugh; how do you guys do it over at the MPC? Do you often relight something in post? (in Harry Potter 2 there was written a special Shake-tool to speed up the rendering by only rendering the normal maps or so, isn’t it?!) I guess that you also do relight stuff by using a normap map and so, and not by importing the whole 3D stuff into you compositor?

Thanks! :wink:
-Gijs

#8

In softwares as the Combustion and Houdini Halo for relight the scene is enough to add and locating the lights (since the images they have a sufficient information for relight). When software doesn´t have this capacity if it uses techniques so that these information (normals, uv and others) are used to illuminate the scene. In the new Fusion 5 it would be possible to work equal to cited softwares?

#9

Dutchman: The reason I mentioned Fusion 5 was cos the original poster did specify it in his post…

I went to a Fusion 5 demo a couple of weeks ago, but I can’t remember if the 3D model they used was animated or not… Because it used the .fbx format, I’d assume it can use animation… I’d be very surprised if it couldn’t.

As for what we use here, it does relighting using normal and position maps. Normal maps are fine when you have a purely directional light, but if you want to do point and spot lights, then you need the position pass…

#10

no animation with the fbx importer… i wonder why they used fbx then??? maybe because its more universal than obj, 3ds etc…

#11

Could you be more specific about how you did this? I just tried the same thing, and it had no effect aside from lighting the image plane. There was no difference if I connected the boolean result to the image plane or the normal image.

#12

Indeed… I don’t get it either… :shrug: I’m really interested in that method, but then first it should work! :wink:

By the way: in an earlier post of me I promised a link to that relight-software that Pixar used. I just found the link back: CLICK!

Have fun! :wink:

#13

po

Hey Hugh!

I have never heard of a position pass before. Could you tell me a little bit about what it is and how it may be generated?

Thanks!
-mental :surprised

/edit: whoops! i found your explanation on VFXTalk but… would you have any idea how a position pass could be created in mentalray for example? It almost sounds like a coverage pass.

The main idea with this kind of way of doing post-lighting is to have two specific passes: Normal and Position

The Normal pass is a render where each pixel in the image gives the information about the normal of the object at that point (basically, which direction the surface of the model is)
The Position pass is a render where each pixel gives the information about the position of the surface at that point.

Between these two, you can generate your own lighting in post…

You will generally need custom written software for whichever application you use to be able to take advantage of this kind of thing. Norman was written specifically for MPC to do just this.

I’ll see if I can find any articles about this…

#14

It should be relatively simply (I’m not a shader writer, and I’ve never used MentalRay, so I can’t help with the specifics)

It’s just got to output a float image where the rgb channels correspond to the world XYZ position of the point that is shown in that pixel.

Obviously transparency is not supported…

#15

Can anyone answer my question above? I can’t get that method to work.

#16

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.