Alpha5: Visible to camera behaviour

[marcatore]

Before opening a bug I'd like to understand how and where use this new feature.
In a previous post where we've discussed about the light visibility, Dade replies that he'll introduce a "visible to camera" option also for the lights so I try it.


When I activate that option, the light (or object) is not visible anymore to the camera, but all the other "channels" are still visible.
This could be useful thinking to the reflection but not really good when shadows still be visible.

Is this correct as behaviour?

If yes, ok but in this way, the shadows reveal the mesh...for example, try to put an area light quite near the ceiling of an interior and de-activate the "visibile to camera", render and you'll see the dark shape of the area light...that shape will blur as much as distant is to the ceiling... ok, a solution could be to leave the area light distant to any other mesh but I think that it's not a great solution for a classic interior rendering scenario... from my POV, this reduce the number of scenario when you should use this feature.

The best should be to have camera visibility, shadows and reflection separated and leave the user free to customize the behaviour.

[Dade]

Is this correct as behaviour?

Yes, it is invisible only to camera rays.

You are seeking a not-intersectable light source like point, spot, etc. I can introduce a flag for making objects not-intersectables (indeed it will be useful only for light emitting objects otherwise it will like not having the object in the scene at all) however the problem with not-intersectable lights is the amount of noise/fireflies they introduce if the scene includes glossy surfaces.
More the surface is glossy-nearly-specular and more fireflies you will have. This can probably be kept in check with variance clamping. I will add the flag but you will see how worse they are than normal intersectable are lights.

[Sharlybg]

Why not simply use area light opacity ?

[Dade]

Why not simply use area light opacity ?

Yes and the advantages and disvantages are exactly the same.

[Sharlybg]

Yes and the advantages and disvantages are exactly the same.

I don't want to be tough but. This can be considered implemented
Btw the way it is really better to consider a 3d scene as a real world project and try to work with real photographer constraints and approach.

[Dade]

I don't want to be tough but. This can be considered implemented

Yes (it is also how camera object invisibility works: they are considered 100% transparent to ray traced from camera).

Btw the way it is really better to consider a 3d scene as a real world project and try to work with real photographer constraints and approach.

Exactly.

[marcatore]

Btw the way it is really better to consider a 3d scene as a real world project and try to work with real photographer constraints and approach.

Exactly.

Ok...I quite agree with you but, again, sometimes is not the quickly and efficient way to render an image (as efficient I'm speaking about time spent on build a "real" 3d world).
Just to explain my POV and avoid to pass as the "always requesting kid".

My daily job is to render stores my colleagues design and in this scenario you have, at least, 2 option about lighting:
- if you have the lighting project (unluckly and very often this will arrive after my renders but we should imagine that I start render only when I have all also the lighting project) you have to model the light projector with a "real" sized light emitter with the right light data. Then duplicate it for the whole store following the project and orient them.
- make a big approximation of the lighting and distribuite some area lights on the ceiling...for a 130sqm store you should put about 30 of area lights.

Considering that "you always need to do this with less time!!!", it's not requested to make a technical check and the relationship from actual lights and area lights is 3 to 1, what do you think it's the quickest method to reach the "press render button" phase?

I hope that this example will not be considered in a rude way . It's just a report for a productive discussion.

[Sharlybg]

My daily job is to render stores my colleagues design and in this scenario you have, at least, 2 option about lighting:
- if you have the lighting project (unluckly and very often this will arrive after my renders but we should imagine that I start render only when I have all also the lighting project) you have to model the light projector with a "real" sized light emitter with the right light data. Then duplicate it for the whole store following the project and orient them.
- make a big approximation of the lighting and distribuite some area lights on the ceiling...for a 130sqm store you should put about 30 of area lights.

I've already encounter such large scale indoor project with a big chineese store to render. It can be challenging but when you have the right approach it is pretty easier and realistics.

1/ Use ALT + D instead of shift+D to duplicate your area light

2/ Then after Alt +D duplication tips light position/distance before hit enter (eg: alt+d---x(axis)---1(1BU or 1Meter)--- Enter)

3/ When it's done leave the duplicated area light selected and automatic replicate the previous operation by holding Shift key and press R key any numbers of time you want your light to be duplicated.

4/ After this you have your first spot light line then you need to cover all the ceilling. Just select your line of spot and auto replicate it by using tips 1&2.

EDIT

5/ for orientation you can use blender builtin "Random transformation" use spacebar search tool to find it.

The Pro of this:

___ AlT + D will link all your area light setting ( changing power/gain/efficiency/color/scale will update all area light)

___ Your render will denoise faster as you have intersectible light

___ You keep all the flexibility and modification are easy to implement

I hope this can help you a bit.

[marcatore]

Thank you Sharlybg,
nice Blender tips, some of them I've already known but nice to see something new.

So, if I model correctly the spotlight, what kind of light should I use? How do you "model" the light distribution? Do you leave that the light reflects into the projector or do you use IES profile?
Have you a sample of something similar you did in the past?

Just screenshot, to understand how you did it in detail.

Thanks in advance.