[ricardo]

I thought that was possible to change in each mxi file interactively to control each lights ... and i think that that it's a great feature ...i did not want afford a critic ...
my question was related to the fact if a complete correct set-up with all the lights will require a time that is the sum of the single passes done with a single pc and not in a network

i know that there will be great advantage...

By whatt I understood, that's exactly the point. Each light generates a light path that is calculated once for each pixel at each pass. So rendering 10 lights will approximatly take the same times as rendering 10 X 1 light, given that they have the similar geometries.

[Joss]

As i understand, now maxwell tracks every light source contribution to a whole solution, and allows to change it later without all these slow computations like billions of rayhit tests. It's a principle of basic relighter - we can imagine it also like rendering scene with every(and only) light source into separate pass, and blending them in realtime in composer. Possibly NL uses some 3d structures, similar to photon maps.

What i'd like to ask is:
1. Is it possible to make quick render from another _arbitrary_ camera, after merging enough mxi's?
2. Is it possible to transfer this data to host application somehow, and change light values from it(and, in turn, rerendering it's with the same speed, we've seen in avi)?
3. Is it possible to implement texture baking, using this technique?

[rivoli]

it's fun, but i can't stop watching it right now, i guess it's because i can make up what the guy who's moving the mouse says. btw, whose pointer is that tom? is that you?

[big K]

really cool stuff ! that´s really a fantastic feature.
(i see me adjusting all this light intensities and having so much fun !)

[Jozvex]

Just wait til we can do per material light contribution too...

Nahh I'm just joking. But seriously though, if you tracked what each material contributed to the scene (sure that's a lot of info) you could possibly tweak shaders in realtime too.

Maybe in Maxwell 2,

[mverta]

Maybe in Maxwell 3.

_Mike

[Jozvex]

Ahh but you see my plan is to get my friend who's got a PhD in biotechnology to clone Oscar a few times so that there's coop Maxwell development.

[jdp]

sweeeeeet!!!! and you guys show that on valentine's day? do you want people around here falling in love with Maxwell again? you succeed!

[gadzooks]

What about the possiablity of changing color of the lights? Say from white to a pale yellow will this update in real time. Very sweet.

[garyswindell]

This is a really neat feature (adjusting lights in real time).

Can't believe it took so many pages for somebody to say this but...

Are all of the features that were announced for V1.0 complete??? If so I know everyone would like to be able to download it. Seems a bit silly to spend a lot of time on this sort of thing when there are serious gaps in basic usability.

Seems like we are suffering from "it's really fun to program neat new things but no fun at all to work on the basic things".

[ricardo]

This is much what I figured out so far:

The MXI file keeps for each pixel an equation like this:

f(L0) + f1(L1) + f2(L2)...

Where LX is any given light and fX is the resulting ligh amount for that light after interacting with all the objects in the secene.

Probably fx returns a light value and a SL value for each light in each pixel.

The calculations go thru some kinfd of pseudo random order, so some pixels hold highier SL values for a given light after each pass.

The seed for this pseudo random system is machine deriverd, something like ethernet address or I don't know what.


The pros:

Changing a single light's intensity returns immediate results, that's great.

This is pretty much a linear system, so you can have different lightning setups combined after rendering, good.

The number crunching depends on the number of lighs involved, so you can split rendering on a per light basis, that's also great.

The pseudo random calculations will result in different SL's for each light in every different computer, so when you mix them, you just use the best SL and many noisy images can combine into a better looking one, good.


The cons:

If we really have pseudo random, rendering the full scene across many computers might not result in a linear gain. What I mean: the same pixel/light combination may have the highiest SL value in more than one MXI, so you have wasted CPU time. Better render in a per light basis.


What won't work:

F/stop: changing f/stop won't result in DOF changes. Better disable it.

Shutter speed: If motion blur is involved, changing it will not result in changing blur, just quit on that one also.

MXI emitters: You can adjust their intensity, but you can't change the map.

Geometry changes: No go.

ps. The "won't work" assumes the MXI is 2D. If not, we then have unbiased baked textures, what I don't belive will come so soon.

[michaelplogue]

This is a really neat feature (adjusting lights in real time).

Can't believe it took so many pages for somebody to say this but...

Are all of the features that were announced for V1.0 complete??? If so I know everyone would like to be able to download it. Seems a bit silly to spend a lot of time on this sort of thing when there are serious gaps in basic usability.

Seems like we are suffering from "it's really fun to program neat new things but no fun at all to work on the basic things".

I was actually thinking the same thing. These features are really terrific, and will really improve workflow. However, not to sound like a wet blanket, but has as much time been spent on those features that are the most needed - as expressed by the members of the forum? It's really great that NL has developed some unprecedented widgets into the program. However, I would much rather have the basics like functional transparency mapping.

We've got some really delicious frosting, but the cake is still pretty dry......

[garyswindell]

Of course now we're the bad guys who are impossible to please.

[b-kandor]

Just about says it all.


michaelplogue Posted: Wed Feb 15, 2006 1:04 am Post subject:

--------------------------------------------------------------------------------

garyswindell wrote:
This is a really neat feature (adjusting lights in real time).

Can't believe it took so many pages for somebody to say this but...

Are all of the features that were announced for V1.0 complete??? If so I know everyone would like to be able to download it. Seems a bit silly to spend a lot of time on this sort of thing when there are serious gaps in basic usability.

Seems like we are suffering from "it's really fun to program neat new things but no fun at all to work on the basic things".


I was actually thinking the same thing. These features are really terrific, and will really improve workflow. However, not to sound like a wet blanket, but has as much time been spent on those features that are the most needed - as expressed by the members of the forum? It's really great that NL has developed some unprecedented widgets into the program. However, I would much rather have the basics like functional transparency mapping.

We've got some really delicious frosting, but the cake is still pretty dry......

Some people insist on focusing on -in this case- "imagined" negatives. You are assuming that the features your concerned about have not yet been addressed. It's like arm-chair programming. Maybe this feature was a necessary milestone on the way to addressing your issues? I don't know - just guessing. In any case this feature is certainly not a widget... it will save untold hours of test rendering for light intensities.

[michaelplogue]

Some people insist on focusing on -in this case- "imagined" negatives. You are assuming that the features your concerned about have not yet been addressed. It's like arm-chair programming. Maybe this feature was a necessary milestone on the way to addressing your issues? I don't know - just guessing. In any case this feature is certainly not a widget... it will save untold hours of test rendering for light intensities.

As I did mention, I think that this will greatly improve workflow. But this is an extra feature. This function, wonderful as it is, was never listed as a feature when they began this project.

Please don't misconstrue my comments as bashing NL. I still think that MR is a fantastic product. However, these additions do not address the basic problems with what we have in our hands now. I can live without being able to adjust lighting on the fly - I have for years. However, I can't live without transparencies - a basic feature in every rendering package on the market. Others need the glass and sunlight problem fixed.

I would just rather see some sort of progress along these lines than something I really wasn't expecting. The concern is that NL may be getting distracted by really cool widgets - something that could be released with version1.xx. If we can get some assurances that advances are being made in what we need, I think we'd be happier for it.

You don't start decorating a house before the foundation has been poured.