Now that is a subject line to make everybody curious, isn' it? ;-)

Well, here is an idea of mine, that I thought could come handy for our
MovieProject (http://www.algonet.se/~jhubert/MovieProject ), or for any
animation.

I was thinking about the POB (Pov Binary Format) by Thomas Baier (his
homepage: http://ourworld.compuserve.com/homepages/thbaier/ ).
To those who don't know of it: He wrote a custom version of POV-Ray that
reads a POV-file as usual and while parsing it, it writes all objects it
finds into a chunk-based binary format, that can be read by other
applications *much* easier than the very complex POV-Ray text format.

Anyway, what I was thinking about was this:

Changing this version of POV-Ray, so that instead of writing an object
description into a binary file, it creates an OpenGL version of this object
and when the whole scene is parsed, the OpenGL objects would be shown,
instead of the rendered picture. Since OpenGL is quite fast, actually the
parsing speed would be the limiting factor. OpenGL is quite cross-platform
too, so such a version could even be made usable on many platforms.

I didn't think about a version that can really do *everything* and all
objects, but more of a version that can do the most important POV-Ray
primitives in OpenGL. No special textures, lighting etc., everything in one
color (or two or three...), so that one would get an OpenGL preview of the
scene.
Since that would be very fast (limited only by the parsing speed), you could
use it very much for animation previews. Animation in realtime!
You have an animated character and want to check it's movements before the
real render, or you have a scene with objects flying around and want to see
how they are synchornized, that sort of stuff.

I have a bit of OpenGL experience from my TreeDesigner project
(http://www.algonet.se/~jhubert/TreeDesigner ), and was thinking about
programming such a "Realtime Previewer", but I don't know how soon I might
find the time.
So I thought I put it up here, maybe somebody else is interested in doing it
faster than I will be able to?

Johannes -:- MovieProject Coordinator

Post a reply to this message

Johannes Hubert wrote in message <6k3cuo$mai$1@oz.aussie.org>...
>Changing this version of POV-Ray, so that instead of writing an object
>description into a binary file, it creates an OpenGL version of this object
>and when the whole scene is parsed, the OpenGL objects would be shown,
>instead of the rendered picture. Since OpenGL is quite fast, actually the
>parsing speed would be the limiting factor. OpenGL is quite cross-platform
>too, so such a version could even be made usable on many platforms.
 <cut etc>

I thing that's a great idea.. I thought of a simular thing for the pov2mdl
filter to have a povB2mdl so that way you could get a wireframe preview of
an object written on pov code.

But don't look at me for the codeing thou.... I haven't the experience, yet.
I'm working on it slowly.

Fran,

Post a reply to this message

Fran & Melissa wrote in message <6k650e$rno$1@oz.aussie.org>...
>But don't look at me for the codeing thou.... I haven't the experience,
yet.
>I'm working on it slowly.


Actually, I have followed up on my own idea, made some changes, and already
begun with a test version. It looks promising!

I will not create a custom version of POV-Ray. The parsing time for each new
frame would be to long anyway, to make it really "real-time".

Instead I found this solution:

I use Thomas' already existing custom version to create POB-files for the
whole animation: One binary POB file for each frame.
I then read those POB files and show them via OpenGL as a real-time
animation, plus the ability to control the animation (slower, faster, back,
etc.) so that one can really study it.
The turnaround time should be quite good:
Make some changes to your POV-Script, then you have a waiting time equal to
the sum of the parsing times for all frames, but after that, you can view
your animation in OpenGL. That is a lot faster than waiting for the render,
and then maken a FLI or whatever out of the frames.

But don't expect too much: The fancy POV-Ray stuff will not be supported: No
Textures, Halos, complicated Objects like Quadrics, Surface of Revolution
etc. I will try to implement some sort of placeholder for unsupported
objects though. No CSGs either, that is: Union/Merge/Intersections simply
show all objects in them, Difference shows only the first or optionally all
too.
It is really thought for quick preview of animations, not for a real view of
the model (you can always render for that, can you ;-)

And: Since NT is my home, and MFC is my hobby, it won't be portable: Windows
only, sorry!

Johannes.

Post a reply to this message

How about doing stuff like this in Java using Java3D?
I should make life quite easy, is getting more and more portable and has
some hardware-accelaration as well (uses OpenGL on Solaris, AFAIK)

Axel

Post a reply to this message

Axel Hecht wrote:
> 
> How about doing stuff like this in Java using Java3D?
> I should make life quite easy, is getting more and more portable and has
> some hardware-accelaration as well (uses OpenGL on Solaris, AFAIK)

If time permits (which is very unlikely at the moment), maybe... Or
maybe somebody else wants to do it?

Johannes.

Post a reply to this message

How are you doing? I'm very interested in the project. Is it possible to
render a single scene and move the camera. I don't have much OpenGL
experience, but as far as I remember that should be quite easy, or am I
wrong?

Regards,

Tobias

Post a reply to this message

Tobias Ritzau wrote in message <35c726f2.0@news.povray.org>...
>How are you doing? I'm very interested in the project. Is it possible to
>render a single scene and move the camera. I don't have much OpenGL
>experience, but as far as I remember that should be quite easy, or am I
>wrong?


Yep, it isn't too hard.
I got as far as to a first version which supports spheres, cones, cylinders,
boxes and torii. My ISP is down at the moment, but once I have access again,
I will upload a page where you can download the stuff if you are interested.
Sourcecode too, if you are interested in further developing it.

Bye,
Johannes.

Post a reply to this message

Hi Johannes!

He he...this is exactly what I have been thinking of lately....a kind of
parser that "quitely" parses the scene and when getting a valid input(
valid code), previews the scene in OpenGL. This would be best to have
built into the official Windows POV-Ray version, and accessed through a
single toggle-button: "Preview mode" and "Trace mode". 



alt+G to parse the scene. The parser only previews when the code is
vaild. Like this:

[I start coding]

sphere
{
   <0,0,0>
   0.5


finished.]
   

   texture{pigment rgb<1,0,0>}
   


t
finished.]

} 


[now the sphere is finished, and the previewer shows a red little
sphere, beautifully shaded in OpenGL! :)]


Now, for example, imagine that I move to the radius value (0.5) with the
arrow keys. When backspacing over the 5 in the 0.5 value, the sphere has

anything anymore. But, since I only deleted the "5" to replace it with a
"3" and change the radius to 0.3( since I could see in the preview that
the sphere was too big! :) ) - the previewer shows a preview again as




Working with the best raytracer available would be a lot faster! :)

- Mikael.

Post a reply to this message

On Thu, 27 Aug 1998 11:54:55 +0200, SA9### [at] idautbhbse 
        <SA9### [at] idautbhbse> wrote:

>alt+G to parse the scene. The parser only previews when the code is
>vaild. Like this:

<description omitted>

So if my sphere had a radius of .3625, would it reparse and rerender as I
deleted the 5, then the 2, then the 6, and finally reparse and decide not to
render when I deleted the 3?  Doesn't that seem like a lot of parsing?
Remember, some people have posted their experiences with scene files that 
take _hours_ to parse.  

I think if there's gonna be an opengl preview, it has to be done when the 
user asks for it, not automatically.

Post a reply to this message

> So if my sphere had a radius of .3625, would it reparse and rerender as
 I
> deleted the 5, then the 2, then the 6, and finally reparse and decide n
ot to
> render when I deleted the 3?  Doesn't that seem like a lot of parsing?
> Remember, some people have posted their experiences with scene files th
at
> take _hours_ to parse.


A lot of parsing, yes, but this mode would probably be of most help when
creating (smaller) scenes from scratch, not when importing mesh files


the scene looks like when coding, much like you can in programs like 3DS
and such.

> 

> I think if there's gonna be an opengl preview, it has to be done when t
he
> user asks for it, not automatically.


toggle-button that switches between "trace mode" (traditional) and
"preview mode".....

The "Mode tree" should maybe look something like this:

	  		      Parser
			        |
  		              /   \
           "Standard mode"  -       - "Preview mode"
					 /     \
			     Automatic -         - Manual 



----------------------------------------------------------------------
Mikael Carneholm

Dep. of Computer Science
University of Boras, Sweden

http://www.studenter.hb.se/~arch

sa9### [at] idautbhbse

Post a reply to this message