[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kpovmodeler-devel
Subject:    Re: Frame System
From:       Andreas Zehender <az () azweb ! de>
Date:       2006-10-05 17:06:06
Message-ID: 45253B7E.8080906 () azweb ! de
[Download RAW message or body]

Hi Miguel,

sorry if you feel offended by my email, that was not my intention.

I added this sentence for good reason to my email:

> If I understand you correctly you export the first state and the last state and a \
> function to interpolate in between? If that's not what you meant, you can ignore my \
> email. 
> 

I think we are stuck at the point because we do not understand each 
other, so I will go some steps backward.

My guidlines for the animation system are:
1) There is a fixed start scene which is modeled in KPovModeler (that's 
the current state)

2) There are two methods how to animate a scene:
a) scripting with a javascript like language (that's my part)
The script will be called before each frame and has as available objects 
the state of the scene of the previous frame and calculates the state of 
the next frame. That will allow to calculate the animation in discrete 
steps, so objects can interact with each other if needed. The example I 
mentioned with the balls in a room that bounce is an example where to 
calculate the scene in discrete steps. You can't describe the position 
of the balls in functions, you have to make collision detection, make 
impact/impulse calculation and so on each frame.
With special scripts you should even be able to implement a constrain 
system, when the scripts have access to all objects and attributes 
(which will be the case)

Please study that example in detail to see what should be changable:
<example>
// clock is a global variable that goes from 0 to 1 during animation
cube=scene.csg(1).cube(2);
// here the rotation depends on the value of the last frame plus some 
calculated offset
cube.rotate(0)[0] = cube.rotate(0)[0]+sin(clock);
blablub = scene.objectForName("somematerial");
// simple attribute setting, but dependent on another objects state of 
the last frame
blablub.someproperty=4*clock+cube.corner1[2];
// here an object disappears between clock 0.1 and 0.3
if(clock > 0.1 && clock <0.3) blablub.exportToPovray=false;
else if(clock >= 0.3) blablub.exportToPovray=true;
</example>

b) animation of attributes with pathes
There are objects for each type of path type: linear, spline, ...
The path is bound to an object and attribute (say corner 1 of a box or a 
translate object) and is changeable with control points in the 3d view. 
The path gets the clock value as parameter and returns the position for 
that frame

3) Of course you have to preview the animation in KPovModeler.
You have the start frame, that's the static scene modeled. The user 
specifies the number of frames and the animation is calculated by 
executing the scripts and evaluating the position of the pathes. One 
frame after another.
You then have a slider and start/stop/pause buttons as if viewing a 
video. You can jump to any frame and go forward and backward.
That animation is valid until the scene is changed. When the scene is 
changed, the start condition is different or even objects can be 
deleted, so the states of the animation may not have the corresponding 
object any more.

4) Rendering the animation
When the user is satisfied he specifies the start and end point of the 
animation to render (maybe he just wants to render a small part because 
he made adjustments to the scene around clock value 0.3), the animation 
is sent to the renderer, saved as single frames or encoded as video. 
Whether sent as single frames or as whole animation doesn't matter for 
the moment.

Do you agree so far?

Regards,
Andreas



List archive and information: https://mail.kde.org/mailman/listinfo/kpovmodeler-devel


[prev in list] [next in list] [prev in thread] [next in thread]