[naxos]

Hello there,

After some years of frustration, i start a technical challenge here, where the smartests Maxwell users are spending time...
That challenge could have some prices, regardless of good results...

The topic of this challenge is to find a trick / workaround to a maxwell limit : RenderToTexture (aka Texture Baking).
For those how don't know, Taxture Baking is a way to pre-render lights / shadows on an object regardless of it's UV maps coordinates,
so we can use this picture as a texture mpa in some 3DRT softwares (Unity, Shiva, Nova, Virtools, Wirefusion, etc...)

Here is an exemple (rendered with Vray) of what i mean :

[naxos]

i've reached that point :

I place some cameras within the scene, then render each with MaxwellRender...
Then i use those pictures in a CameraMapping method for UV / projection...
Then i simply use default Scanline render with the 3dsmax's feature RenderToTexture...

It works nice with an empty room... but as i put a colomn, a stand, an object... no way except using Photoshop to clean each picture by hand...

Anyone ready to help me on this issue ?

Challenge accepted ?

[Aji Enrico]

The hard part is that Maxwell is not view independent, so some effects will not be shown like clear reflections, refractions, anisotropy, coating, fresnel just to name a few
What can be show is, with little or no quality loss, is "radiosity" type renders like rough materials that doesn't show much view dependent effects.

But what Maxwell could be used for in games or in other real time applications, is lightmapping.
Render lightmaps from Maxwell would be awesome imho, and let the other app (like Unity) care for the view dependent effects like reflections, SSS, refractions etc. in conjunction with the lightmaps.

Maybe I'm wrong here?

[naxos]

...
But what Maxwell could be used for in games or in other real time applications, is lightmapping.
Render lightmaps from Maxwell would be awesome imho, and let the other app (like Unity) care for the view dependent effects like reflections, SSS, refractions etc. in conjunction with the lightmaps.

Maybe I'm wrong here?

That is what i mean...
Reflections and refractions are not intented to be used in real time virtualtour... (some RT engines can handle fake / simili-real ones)...
But just the lightmap... that would be soooo better than biased engines...

There may be a real technical limitation i guess... because even iRay (now full part of 3dsmax) is also unbiased, and also not compatible with RenderToTexture
...

[Aji Enrico]

i've reached that point :
It works nice with an empty room... but as i put a colomn, a stand, an object... no way except using Photoshop to clean each picture by hand...

Anyone ready to help me on this issue ?

Have you tried "visible to camera" settings?
Object Properties -> Rendering Control -> Uncheck Visible to Camera
Heres a quick test with columns and... stuff...

invisible.jpg

visible.jpg

[mashium123]

that's also what i would think.
working with the camera properties helps here.
so with this method, there may definitely be cases, in which lightmap-ish results could be produced.

a very simple example here, using the methods described above: (rendered in ~ a second per frame out of cinema4d)
http://www.mesutcapkin.com/temp/maxwell ... nima_a.mp4

in those simple cases, this is done pretty quick an straight forward... everything's flat and easy...
the more complicated and "organic" the geometry becomes, the more complicated and time-consuming the whole texturing/mapping procedure will become.
but still: with tools like bodypaint, some work in that direction is certainly do-able, imho.

[Michael Betke]

For realtime a good render to texture with mentalray is more an enought. you even can set additional light properties and material in your target real-time tool.

[Bubbaloo]

For realtime a good render to texture with mentalray is more an enought.

For who? Not Maxwell users. When someone comes along and wants to push Maxwell beyond it's current capabilities, suggesting another render engine seems counterproductive.

[naxos]

Have you tried "visible to camera" settings?

I did, and this method is good for empty rooms... so quite good for museum / galeries...
But not enough for any object (column, stand, etc...)

[naxos]

For realtime a good render to texture with mentalray is more an enought. you even can set additional light properties and material in your target real-time tool.

Not agree...

For the quality i seek for, nor MentalRay, nor Vray are good enough... sorry.

[naxos]

Someone at NextLimit answered me about that limitation :

----
Next Limit
Hello Claudio Gallego,

The reason why unbiased renderers can not bake light to texture is because there is no way for doing this in a physically correct way.
On regular biased renderers, you can bake to texture the diffuse lighting (as it is not dependant on the point of view), while the specular highlights and reflections are not baked (because they depend on the position of the camera).
Instead, for unbiased renderers there is no separation between diffuse and reflected light. All light is reflected in fact (and this is like that in the real world).
So this is the ultimate reason why there is not possible to bake lighting in unbiased renders.
Anyway, we are working hard to make Maxwell Render faster...
----

So there IS a real reason (i mean technical, not political / marketing) to that limitation...

An unbiased engine can't render only diffuse light (as needed for lightmap)...

BUT, setting all objects as 242-white & 99% rough shader, we could render a pseudo-lightmap to be used in RT vir-tours...

Or maybe a camera plugin to get each poly regardless of its' UVcoordinates...

Here is my answer :
-----
Faster is great job... and CPU can be ok vs GPU on some renders, Fire is a great example...
I'm just quite sad that each time my customers ask for a virtual tour in their archi, they can see a large gap between still images, animations and real time...
The fact is that if maxwell was compatible with RenderToTexture (3dsmax's texture baking), of course we would only use 99% rough 242 white dull shaders, the idea is to get nice lights / shadows, not all light spectrum of effects...
I'm not sure how the RenderToTexture works, it renders the picture on each polygon as if the camera was moved strait in front of it regardless of UV map...
So i can't put out of my head the idea that it could be possible, even if not 100% accurate, to maybe code a camera plugin that acts like this... a bit like 360° virtual cameras plugins (Vray has one)... or special anamorphic cameras...
Maybe i should think about actually script the camera placement in front of the polygon, uncheck visibility of all-but-this object, render it, and place it at the right UVmap position...
Problem is : i'm not a coder ;-(
anyway, i'll be to Vancouver for Siggraph (my first time, after more than 20 years 3D making), be sure i'll come to see your team there, and ask again for texture baking solutions
-----

So the topic is still alive... just because i can see 2 possible solutions to go there...
- special camera plugin : but i'm not a coder;
- let us set the scene to dull white shaders and rendertotexture...

[Michael Betke]

ral-time engines are always faking all kind of lighing and other effects. so you could have the light physically correct but not specularity, cubemaps and so on. where is the sense then for only having unbiased lightmaps?

secondly most clients which demand real-time viz use it for marketing or decision making. which client or their customers will see a difference between vray, maxwell, mentalray baked lighting running in the real-time engine where other effects still need to be faked or looking out of place then? there will be only a very small group of endusers which are willing to pay for the additional effort.

in my opinion its important to build a good complete package using several elements to make it believable. not looking unbiased at all costs.

[naxos]

which client or their customers will see a difference between vray, maxwell, mentalray baked lighting running in the real-time engine

You almost right, except the light with maxwell is soooo far better than Vray / mentalRay...

anyone, even novice can see the difference...

[AndreD]

fryrender has a built in walk-through feature and it is free for existing fryrender users...
(at least, this has been promised.. )

[naxos]

fryrender has a built in walk-through feature and it is free for existing fryrender users...
(at least, this has been promised.. )

Some times ago i bought Fry, specially for that promise...

But it turned into a pseudo realtime renderer, just like Fire / VrayRT... called Arion...

The walk-through feature is not ready yet....