first official try

[djegoo]

yo
so hello, i m Diego, i m a brand new user of indigo, and i think i already creted this post but something went wrong when posting

so, i m architect, and i use Sketchup for 3D, and i ve been travelling through many renderers and i wanted to show my first work. the scene is not exceptionnal it is not the point, it was just a test on technical stuff to see how improve speed and accuracy when rendering scenes with natural and artificial lights, direct and indirect lights etc, many objetcs creating shadows, etc... because when rendering architectural projetcs, scenes are often complicated.


so i tryed the followings, and i would like you to tell me if i did the right stuff or if i could change something il my way of using indigo!

-Tracing method : Bidir Mlt, the guide says that for interior scenes with lots of indirect light and caustics it is better. am i right?
-Exit portals in front of the window
-Bump on the ground 0.01, to not make it too weird.

There is an issue on the lights but it comes from the 3D model from the sketch library. ther is just one surface.

hope you can tell me if i m using it the right way, and if there are some things i can improve im my technical use.

Regards. Diego

this is a 3hours rendering on a 2x2.4Ghz and 2Go memory.it is not finished yet, some reflections and gloss are not yet totaly done. i guess it takes more or less the same time as mawxell or other unibased renders with same config computer?
i wanted to mention that i really appréciate indigo because it is SO EASY to use, and so Sketchup friendly integrated!!

[Pibuz]

... because when rendering architectural projetcs, scenes are often complicated.

DON'T TELL ME!

hi Diego, I'm an architect too, and I'd really like to see where you will bring SU/Indigo in your projects!

Before I comment the image: which version of Indigo/SkIndigo do you use?

[djegoo]

Hi Pibuz
THanks a lot for answering

So, First of All I apologize because my post appeared twice !!! i did not see that uhhh, it needed to be approved until i got the mail. so i think the others who answered in the other post will continue here.

so Thank you ZomB for the pure black answer. very helpful. another question though : in the advanced tutorial, the exit portal is black too, but not pure black. is ot for the same reason?

so, i m a SU6 user, and i use indigo 1.1.18 and skindigo 1.1.16.

Sure pibuz. Actually i have not yet done a lot of images on Indigo, but i have some spares i will share soon.

to answer From the answer post, ok i will wait till my renders get 2000 sample almost. actually i don t have a supacalculator computers and many funds for renderfarms, but i have patience. i am resuming the render to see the progression. i will post it soon.

[CTZn]

in the advanced tutorial, the exit portal is black too, but not pure black. is ot for the same reason?

Generally speaking no, not at all. Indigo only uses the geometric data from exit portals, their shaders are strictly ignored.

Welcome djegoo !

[djegoo]

ok thanks a lot
i will post the progression of this stuff, it is at 7 hours now, it is quite clean, but it seems that the reflections and details need to bake more and more.

oh by the way. if for example i have a yellow, and a Blue and a orange light, does it matter on the render speed or is the main factor the difference of power/luminisity?

[djegoo]

hi

i am back, i have a new version of the same image, but with new questions

this time, i rendered the model on "camera" settings instead of "reinhard", and it seems that it went faster, even on the result, and in the number of sample/pixel.

with the camera setting, in 9hours, it reached 1200sample, although with the reinhard, in 8 hours, it reached 800 sample (see proportionnally camera seems to be faster)
so my question is : is it just luck? or is there a particuliar reason? i am very interested.
and visually, the result is really more close to ta final result!

for this version here are the corrections :
-Reduced the red saturation
-Reduced bump to 0.015
-Added some indigo shiny plastic material to the fan of the ceiling
-added some thickness to the lamps
-Changed the light bulbs to pure black

i also guess that these corrections helped a lot to improve speed, but i am still intrigated by the reinhard/camera story. i will test a render of the same image with corrections with reinhard to see again because i haven t tested with reinhard with the correction

by the way. are there types of scene where i should prefer reinhard/camera? or can i always use the same?
sorry for all these questions, i serached a bit on the forum but there are points i haven t found

regards
Diego

[Pibuz]

Hi Diego!

As far as I know, camera and reinhard tonemapping shouldn't trigger different rendering times, or image quality. The main difference (at least in skindigo) is that reinhard is auto-exposure-ing (): my first tests were actually carried out through this tonemapping, because I found that one didn't have to waste time setting the camera parameters.

Now I actually use reinhard rarely, because I find that it produces somehow flat images, if compared to the ones you have using the camera tonemapping. So I suggest you do your tests with reinhard, and you choose Camera for your final render.

Another thing I have to say is that when I used reinhard, there was no "multilight" feature: if you didn't get the right exposure, you had to stop the rendering and re-lauch. THAT was a loss of time
Now that we have multilight, you can set the camera wrong, but you can adjust the settings while rendering, and obtain a correctly-exposed image as well. So I think there is no longer a strong valid reason to use reinhard as a tonemapping, after all.

A tip about the bump: it is a very sensible parameter; to speed up cthe calculation and achieve more realistic results, don't use SUCH great values, like 0.01 (1cm). I've been using SkIndigo since about three years now, and I NEVER found a case in which more that 0.009value was needed, and that was a very strong value

The mat applied to the fan makes it look more realistic, but be careful: you see those dark edges? They come out because you applied a phong material to a too-lowpoly object. When you apply a phong or metal shader, do some test before launching the final render, and see if the object is detailed enough not to make the dark blotches on the border appear.

That's all for now

Keep testing!

[djegoo]

hi Pibuz

thanks for these tips. Yes i agree the reinhard has something flat, like there slightly not enough contrast, and the whole scene becomes totally visible, altough sometime there should be some parts darker than others, or colors appearing not "bright" enough.

I think i ll keep on camera, it is not complicated to set a camera and i find the basic setting correctly setted for quite all type of scenes!

for the bump, i was nicely surprised, because i finally prefer this version of the floor, it looks less like a usured parkex! it is just that in the tutorial, it said something like "never use more than 0.02 or 0.03" so i wanted to try the effect

yes i see the black edges on the fan i also read something about issues of low polys on the forum.

ok now i have a final question, i think it will be the last for this test, and after i will be able to set my lights correctly!
for the lights, i created a bulb in each lamp, and i attributed the light material in the "assign preset"
i read somewhere that the power was shared through the surface of the material, so i multiplied by nine the initial 34Watt value (since i have 9 bulbs), but it did not light anything so i tested and tested again untill i had these settings (these are the ones of the posted iamges)

power 180000
efficacy 100

instead of the basic
power 34
efficacy 50

i obtained the values by trying many times but actually, if i divide 180000 by nine, does it mean that each bulb i added in the light fixtures have 20000 watts?? or is it not in watts (i thought it is in watt because in the preview scenes, the appearing name is "34w" and the power is 34 too, and other lights name also correspond to power :p tell me if i m wrong).

[Pibuz]

I've never understood what was the physical limit of the representation of lights. I mean: sometimes a single small room is lit only with a 20W bulb.

You think: how can I have a bright image with only a small bulb with a small power? The answer is: you gotta play with camera parameters, OR change the power. It's clear that in both cases you will brighten up your render.

The point is: when have I to play with lights' powers and when instead it is better to adjust the camera settings? The answer this time is: you choose what it is convenient for you. I don't have to say that it is somehow ridiculous to set a bulb to be 100.000W strong, but could be a solution if you don't want to play with ISO and shutters.

As a general rule, it is always a GOOD thing to model the light sources at their real scale, and to assign them reasonable (as in real life) values. This helps especially when you get to have MANY emitters in your scene, so the powers are well-balanced, so none of them prevails over the others.

When you have one single emitter, then I think: adjust things as you want
You can't be wrong

[djegoo]

yes very interesting! changing the iso and the camera parameters helped me but i had to turn the sun off. i tryed for example in a "sketchup background color" and checked black, and playing with iso, speed and aperture worked.
in this scene, my problem was that the light emitted by the 9 x 34w bulbs were not visible because of the sun, whatever the camera setting was

i guess it is because the Sunlight power value is something unproportionnal to little bulbs??

[Pibuz]

exactly

[djegoo]

ok! thanks ! i ll feel less ashamed to put my 20000 watts on a little light spot then !

thank you for all your answers! i ll try to render some models i have correctluy now!

[djegoo]

i have another question.
I tryed to understant the supersample factor story by searching on the forum and on google, but i a m not sure i totally understood.

Did you set the super sample to 1 ?
This is very important parameter for doing large renderings. The SS number multiplies your image in memory. So when set to 2 you are actually rendering 6000 pixels instead of 3000.

btw: How much memory is "far more" and are you using 64 bit op sys?

http://www.indigorenderer.com/forum/pos ... 17&p=74702

this comes from here.

So, when i render, it writes in the log "supersample factor set to 2... etc....etc..."
if for example, i render a 640x480 image, according to what is quoted above, it is in reality a 1280x960?
my second and third question are then : if i set it to 1, will it reduce the image memory size? so i can increae the image resolution?

[PureSpider]

So, when i render, it writes in the log "supersample factor set to 2... etc....etc..."
if for example, i render a 640x480 image, according to what is quoted above, it is in reality a 1280x960?
my second and third question are then : if i set it to 1, will it reduce the image memory size? so i can increae the image resolution?

Yes it does, this is to prevent some ugly alaising around bright spots in the image by rendering it double the size and then scaling it down.
And yes if you set it to 1 you can render images with double the resolution with the same amount of memory used.

[djegoo]

thank you very much. it is exactly what i wanted to know

then, i have a last question : i am testing right now something :
I have an iamge, i render with the SS factor to 2, and then, the same image but the SSfactor to 1
According to what we talked about, the second one should render faster? (but in lower quality in some way?)
this is what is happening on my computer right now, i wanted to know if it was some kind of coincidence or if it was what should theorically happen

(sorry i ask many question but i like this renderer and i would like to know how to find a compromise between quality and speed! i m so impatient to see my models in a realistic way!)

EDIT: also, i tryed the Skindigo2 with the indigo2, and it seemed to be faster to render than my indigo 1.1.18 /skindigo 1.1.16, although in the render log, it did less sample/pixel (and also sample/second)
Is this possible? actually i don t know a lot about it, so i guess smple/second shows the overal speed while the Sample is the overall progression.