Maxwell Render | Official public forum

This might be plain stupid, but I'm asking anyway. Would, in theory, a built in grayscale rendering mode (no post processing), decrease rendering times?

No, because Maxwell is not RGB based. In spectral form grey is just like any other color.

ah, I see. So that would just be restricting it to one wavelength of light?

How about, a greyscale channel, where all the light would be filtered neutral white, and all the materials would be overridden with grayscale reflectance, but keep their layered structures, greyscale transparency, roughness and such - would such a render, render more quickly?

That is, If I did all these changes manually to my scene, would it render quicker? perhaps it's the same question, I'm not sure, but just to confirm,.

bjorn.syse wrote:ah, I see. So that would just be restricting it to one wavelength of light?

No. In spectral form, grey is a shape of curve just like red or blue are others and it's described in the range of visible spectrum. In spectral space, there are no channels or multiple layers forming the final color. Sensor is the Maxwell filmback. The computed spectral result is finally being fit into 32/16 or 8 bit RGB gamut according to selected color space.

For example, below data has remarkable red, green and blue peaks so, it's a good example showing how you should imagine the grey. Conversion to RGB is something like separating this data with red, green and blue filters. This means, something grey to you might be something yellow to another creature who has less blue perception in its retina.

Don´t know where to ask this so this topic seems closest.

I know Maxwell internally doesn´t work in RGB and can target most professional color standards, However is it possible to output 30 bits colour straight to your monitor (Nec PA 10 bits panel,30 bits or higher video card and windows 7 all should support this). The idea being of course that the user can make the decision visually where to go with the rendered image?

I´m considering that Nec for other applications which seem to support this so what´s Maxwell doing? Sorry if this question is already answered. I couldn´t find it with several searches and I have been away from Maxwell for years.

tom wrote:

bjorn.syse wrote:ah, I see. So that would just be restricting it to one wavelength of light?

No. In spectral form, grey is a shape of curve just like red or blue are others and it's described in the range of visible spectrum. In spectral space, there are no channels or multiple layers forming the final color. Sensor is the Maxwell filmback. The computed spectral result is finally being fit into 32/16 or 8 bit RGB gamut according to selected color space.

For example, below data has remarkable red, green and blue peaks so, it's a good example showing how you should imagine the grey. Conversion to RGB is something like separating this data with red, green and blue filters. This means, something grey to you might be something yellow to another creature who has less blue perception in its retina.

Does maxwell have a continuous curve like that? Or does Maxwell turn RGB into 9 or 12 or more spectral samples and then back again to RGB for display?

Is there something I can read regarding the mathematics of converting and RGB value to spectral samples, or is that proprietary for Maxwell?

I think we need Tom for both our questions. Maybe he is on holiday?

max3d wrote:However is it possible to output 30 bits colour straight to your monitor (Nec PA 10 bits panel,30 bits or higher video card and windows 7 all should support this). The idea being of course that the user can make the decision visually where to go with the rendered image? I´m considering that Nec for other applications which seem to support this so what´s Maxwell doing? Sorry if this question is already answered. I couldn´t find it with several searches and I have been away from Maxwell for years.

Maxwell uses Nokia QT cossplatform UI framework, so it's limited with its features when it comes to display. Currently we can't support simulation of color profiles on monitor but, you can still save 32 bits EXR or TIFF and display it using another imaging application.

jfrancis wrote:Does maxwell have a continuous curve like that? Or does Maxwell turn RGB into 9 or 12 or more spectral samples and then back again to RGB for display?
Is there something I can read regarding the mathematics of converting and RGB value to spectral samples, or is that proprietary for Maxwell?

Yes, Maxwell has a continuous curve like that. It's not dealing with RGB during the render in any way until it comes to the tonemapping process. We can't share formulas, sorry.

tom wrote:

max3d wrote:However is it possible to output 30 bits colour straight to your monitor (Nec PA 10 bits panel,30 bits or higher video card and windows 7 all should support this). The idea being of course that the user can make the decision visually where to go with the rendered image? I´m considering that Nec for other applications which seem to support this so what´s Maxwell doing? Sorry if this question is already answered. I couldn´t find it with several searches and I have been away from Maxwell for years.

Maxwell uses Nokia QT cossplatform UI framework, so it's limited with its features when it comes to display. Currently we can't support simulation of color profiles on monitor but, you can still save 32 bits EXR or TIFF and display it using another imaging application.

jfrancis wrote:Does maxwell have a continuous curve like that? Or does Maxwell turn RGB into 9 or 12 or more spectral samples and then back again to RGB for display?
Is there something I can read regarding the mathematics of converting and RGB value to spectral samples, or is that proprietary for Maxwell?

Yes, Maxwell has a continuous curve like that. It's not dealing with RGB during the render in any way until it comes to the tonemapping process. We can't share formulas, sorry.

Hi Tom, so you have a small bot which crawls the forum looking for your name

I was not aware that QT had this restriction. Seems strange to me that the promoters of Deep Color forgot all about it in their C++ framework.
Anyway, it´s said to hear as we now have the blistering fast previews and I would like to what materials do in different color spaces before rendering the scene out. The NEC can switch on the fly between sRGB, softproofing and video standards which would be very convenient to see in the early stages.

Is this a wishlist item to get a separatie preview renderer output window with 10 bits res? It could be implement while sticking to the QT environment I think.

QT ... the promoters of Deep Color.

Just curious, from where did you get this idea?

tom wrote:We can't share formulas, sorry.

Are there any resources you can suggest that discuss conversion between RGB and multispectral samples and back again that don't involve Maxwell proprietary info?

I'm trying to figure out if I can fake some samples of a color photo in Photoshop using evenly spaced Select > Color Range masks, or maybe by using the other colors (yellow, cyan, magenta) of the Black and White adjustment layer so that I have, say, 6 color channels representing an image, and a way to combine them back into something close to the original image.

Is there some 6 x 3 or 3 x 6 matrix that would convert between RGB and RYGCBM and back?

Is magenta even a spectral sample, since it is where the rainbow 'warps around again on itself?'

Just to mess around.

JDHill wrote:

QT ... the promoters of Deep Color.

Just curious, from where did you get this idea?

That came out my cold, sour throat and general misery in an environment which is about 30 degrees which together makes my brain malfunction. The Q must have reminded me of Quadro´s and after that completely silly association (I was looking for a Quadro as my GTX460 is rendering Octane tests) it was a small jump to Deep color. At least a small jump but a bridge to far for mankind

Would still like to have a independent viewer window in 30 bits integrated in studio as an option though. I understand that this requires a new tone mapping for this wider gamut, but I think I wouldn´t be the only one loving to have it. I hope if I reread this wish tomorrow I won´t discover I slipped up again....

jfrancis wrote:

tom wrote:We can't share formulas, sorry.

Are there any resources you can suggest that discuss conversion between RGB and multispectral samples and back again that don't involve Maxwell proprietary info?

I'm trying to figure out if I can fake some samples of a color photo in Photoshop using evenly spaced Select > Color Range masks, or maybe by using the other colors (yellow, cyan, magenta) of the Black and White adjustment layer so that I have, say, 6 color channels representing an image, and a way to combine them back into something close to the original image.

Is there some 6 x 3 or 3 x 6 matrix that would convert between RGB and RYGCBM and back?

Is magenta even a spectral sample, since it is where the rainbow 'warps around again on itself?'

Just to mess around.

Do you have access to the ACM papers? There is a lot of information about the back and fro conversion / tone mapping algorithms. Maxwell as I understood is by the way capable of 3d luts so could build very precise profiles to convey exactly what the intended color space should be on your monitor. In general I think it would be easier to use a 30 bits color space for spectral to RGB conversions.

seriously?

who gives a rat's ass (arse for the uk folk) how it works...... just work 'with' it....

rusteberg wrote:seriously?

who gives a rat's ass (arse for the uk folk) how it works...... just work 'with' it....

Maxwell is about physical reality. Internally it works very precise, but you can´t see it on your monitor unless you first save it and reopen it in another app. Wouldn´t you like to have instant feedback on your materials and ligting setup? You don´t have to be a tech nerd for that wish.
Look at how many people model and render a car or furniture for a living. How can they be sure the chairs match the paint if they have to work in sRGB. Exactly nobody can as it will depend on the light sources in the scene.

In your showroom lights setup it looks matching, but exactly the same model, materials etc will look completely different in daylight and worse your curtains no longer match with your colored glass table. How much time do you you lose this way? You can´t see a difference in sRGB but what if people take the output to priint, video, or wide gamut aRGb.

I for one would like to see that in my new preview render as correct as possible. It´s a cyclus of corrections, test renders, etc which could become much faster if all the 16 bits were output to your monitor. This only works for people with large LUT´s on board and preferably a 10 or more depth screen, but hey you can buy them under 1000$. So it´s not an academic excercise but a real world problem.

Hope this explains it in laymen´s terms. Google will no doubt tell you more with ´spectral rendering´+ metameric + RGB as a search instruction.

Max