Page 1 of 2

WinOSI ?

Posted: Sun Oct 20, 2002 9:01 pm
by rosss
Just wanted to say that their is a great freeware SMALL renderer ..

http://www.winosi.onlinehome.de/

Re: WinOSI ?

Posted: Mon Oct 21, 2002 2:09 pm
by Debaser
Looks good indeed, and it's opensource :!: could be very useful.
But there's not a lot of documentation yet.

Posted: Mon Oct 21, 2002 11:42 pm
by kaktuswasser
looks good,but seems to be very slow :?

Posted: Tue Oct 22, 2002 3:49 am
by MrMunkily
visual quality aint that great neither. lots of grain from what i saw. not an artists render, rather a true simulation of what light would do. not exactly what an artists program needs to be... but interesting nonetheless

Posted: Tue Oct 22, 2002 9:07 am
by protoplazm
it may have been my computer, but the speed of WinOSI is TERRIBLE...

but i checked the gallery online, and the results look pretty good. i don't know if the speed is acceptable though

Posted: Wed Oct 23, 2002 12:59 am
by MrMunkily
it has no optimizations and no appoximations. that's why it looks so bad unless you leave it on for 100 days. those images took like 5 days to render.

Posted: Wed Oct 23, 2002 9:36 am
by protoplazm
that explains a lot :p

Posted: Thu Oct 24, 2002 2:22 am
by Noiprox
Do not judge WinOSI before you even know what it is!

It is an experiment in physically-based modelling of light. It is not intended for artistic use. It does represent what will probably be standard for CG eventually. By modelling the actual photons as rays, it is possible to simulate all light phenomena with perfect accuracy. Unfortunately, with current computing power, that requires days of calculation. There was a time but a decade ago when a typical frame of a typical game would be all but impossible to achieve, even pre-rendered. Now it's realtime...

WinOSI is very impressive in that it models light with amazing physical accuracy. For Blender, it would be better to look to Renderman, raytracing and the like instead.

Posted: Wed Nov 27, 2002 8:02 am
by MrMunkily
sure. it's not really a renderer.

It's not realistic at this point.

I have yet to see a GI solution that was actually completed. everything looks half caluclated with a lot of grain.

Posted: Wed Nov 27, 2002 4:18 pm
by green
Noiprox wrote:Do not judge WinOSI before you even know what it is!

It is an experiment in physically-based modelling of light. It is not intended for artistic use. It does represent what will probably be standard for CG eventually. By modelling the actual photons as rays, it is possible to simulate all light phenomena with perfect accuracy. Unfortunately, with current computing power, that requires days of calculation. There was a time but a decade ago when a typical frame of a typical game would be all but impossible to achieve, even pre-rendered. Now it's realtime...

WinOSI is very impressive in that it models light with amazing physical accuracy. For Blender, it would be better to look to Renderman, raytracing and the like instead.
I dont see alot of checkered floors and glass spheres in proffesional photo's...

Posted: Fri Nov 29, 2002 2:55 am
by Noiprox
Green, you managed to mispunctuate and misspell in a single chunk of destructive sarcasm.

If you had read what I wrote more carefully, you would notice that I never called it photorealistic. It is more physically accurate than other algorithms. Those checkered floors and glass spheres you refer to are clearly due to the limitations of the scene description language supported by WinOSI, not the rendering algorithm itself. It is theoretically possible to render a scene photorealistically with an approach like WinOSI's, but not with raytracing, and definately not with REYES or scanline algorithms.

MrMunkily, if you think WinOSI is not a renderer, you should look up the definition of the word. WinOSI is most definately a renderer, just not one designed for the purposes that you people all seem to assume it was intended for. WinOSI is more of a scientific experiment than anything else.

As for Global Illumination, do you mean WinOSI's examples are not complete GI solutions, or do you mean GI in general? If you knew anything about the physics of light, you would know that light levels on surfaces are actually defined as the limits of infinite photonic reflections. There is no such things as calculating a GI scene to "completion". By modelling the actual photons, it is possible to generate pixel-perfect images, though, and those will be "complete" in your sense of the word, but that requires a fair bit of supercomputer time for even a simple scene.

There are actually scenes that have been calculated to extremely high degrees of accuracy, the classical one being the Cornell Box, of course.

I am not defending WinOSI for the sake of WinOSI.
I stand for truth.

Posted: Fri Nov 29, 2002 5:10 pm
by neil
This is why Blender needs besides any integrated renderer(s) the ability to easily interface with external renderers. No-one can agree what the ideal renderer is because everyone has different requirements/interests/patience (in terms of rendering times). Personally I think once he has made the upgrade to handle complex geometry better this will be a very good option to have along with other renderers.

Neil

Posted: Sat Nov 30, 2002 5:59 pm
by green
""""""""""""""
It is theoretically possible to render a scene photorealistically with an approach like WinOSI's, but not with raytracing, and definately not with REYES or scanline algorithms.
"""""""""""

I think what you meant to say that it is impossible for YOU to do it.
The artists doesnt blame the tool when he fails.. he learns new ways to get to the goal.

People like you just want a MakeNiceLookingImage button, *sheesh*

Posted: Mon Dec 02, 2002 12:42 am
by alltaken
the renderer doesn't look to bad.

the images are grainy because the camera is picking up only photons that reach it (because "the renderer is so advanced" )(LOL)

na i think it looks nice and the images are rather real looking. (even if they look a bit washed out)

Hmmmmm
please don't start a fight :cry: :cry: :cry: :cry:

it would make me sad.

wait for a new version i say, it still has a few bugs.

The Problem...

Posted: Mon Dec 02, 2002 8:08 am
by kniffo
Well, beside that there's a problem. Not all of the renderers, WinOSI for instance or Light++, can take renderman rib-files as their input. Light++ takes C++ text. vmd-files are of another format than the others. So, I don't want to talk long about this - what do you think of desinging an interface definition language, so that we would have a bison for blender. The designer of an interface to any renderer would provide an interface definition to his format and the blender would then use this definition to produce the right format.

Would'nt that be nice? Well, if this is a good idea, I would start such a thing but I need your concerns! :)

/Kniffo