Introducing myself and my program FractaloScop

[gurroa]

Hi there,

I am programmer from Czech, EU.

Two years ago I was confronted with a challenge to create a software that will be presenting animated fractals in real time as a tool for VJs. So I have programmed the FractaloScop. This program is used the same way as a kaleidoscope but instead of glass and light you play with fractals. It runs on DirectX HLSL and the base is programmed in Delphi. Now it even reacts to sound levels from mic and has a powerful animation mixing tool implemented that allows user to mix live animations the same way you create music in Fruityloop or similar software.

While in the process I've found out way how to export hi-resolution pictures. Currently I ended up at the resolution of 2.4Gpix resized into 37Mpix final picture - square 6.144 - but rendered into 12.288 * 16xAntialiasing. I'm displaying and selling those pictures occasionally as a 50 cm to 1 meter squared wall picture. And why is the render so huge? Because it look a lot more smoother once printed - even a lot more than the 16xA itself - you can imagine it as a 256xA - but I have tried that too, render time was amazingly slow but results somehow lower than pure resizing. And as those fractals are not so cool as most seen here their main usage is live animation - readable shapes and low complexity is needed.

This year I have implemented 3D fractals. With live (real-time >25fps) flying mode now with up to 1000x zooming ability for MandelBulb and 100x for MandelBox (too much complexity to render smoothly). First I was using this flying mode just to find a perfect position for rendering photos and recording videos. But now I am using it as the Amusement & Interactive installation on Psy-trance festivals - and I called it Fractalium. Visitors are controlling the fly using two old-fashion joysticks. Because I was told that it is not so common to see a live fractal flyer I thought it is the right time for me to enter world-wide fractal community. Also I would like to share and discuss some features with people who were working on those problems for ages. Especially I am interested in texturing and colored-light mixing within DirectX HLSL.

You can visit my facebook page where I post the most (only in czech language but pictures are international, right?):
http://www.facebook.com/gurroa.fractals

You can visit my quick gallery view - beware the page contains more than 60MB of jpeg preview images (1024x1024) - no thumbnails included, not good for a low connection:
http://www.gurroa.com/Fraktal

You can visit my youtube list with various animated 2D fractals and few latest 3D fly-throughs:
https://www.youtube.com/playlist?list=PLZ5lrZ_eoY86LGNjJmc1FOXqa8C9zQkut

As you will probably find out I am only using few types of fractals - Mandelbrot and Julia (3 variations) and Duck fractal - but I have added some modifying features to be able to animate them so those are quite enough and not always recognizable by their origin.

I have yet implemented only single MandelBulb and single MandelBox. I had to spend most of the time on balancing latest version of my version of ray-tracing algorithm since I had found out the zoom ability and had to rewrite it massively.
Also I can now modify geometry the same way I'm doing in 2D. I think that those two starting points will be probably enough for a time too. 1000x zoom means that the MandelBulb will grow 1000 times in every way. And I must say that 5 iterations on those relatively deep details creates quite a horrible amount of mass to explore. Not limitless but something like a small planet 60 km in a diameter. And I think it can get even a lot larger.

My deepest recorded glance is from 25x900 zoom (22.500x) with curved geometry - you can watch it on youtube. This deepness is still waiting for a live flying or perhaps current generation of graphical card (still using older GTX700).
<a href="https://www.youtube.com/v/A230Gww_vbo&rel=1&fs=1&hd=1" target="_blank">https://www.youtube.com/v/A230Gww_vbo&rel=1&fs=1&hd=1</a>

[cKleinhuis]

hi there and hello and welcome to the forums, i write small vj plugins using shader code as well under the label of spack-o-mat plugins for vj software resolume, too bad there is no sound on the video

[tit_toinou]

Welcome !

Nice to see people VJing here . Never seen a live Mandelbulb in a psy trance festival in France .

While in the process I've found out way how to export hi-resolution pictures

Is it like fragmentarium ? Divind the image in little blocks ?

[gurroa]

Is it like fragmentarium ? Divind the image in little blocks ?

Yes. I suppose.

Nowadays I am rendering those huge images in small frame sized only 128x128 pixels.
All of mine fractals hlsl has at least two loops evolved.
And rendering smaller parts give you a lot faster result - comparing to larger part.

I have done some tests and find out that 128x128 is the average size possible.
Smaller is rendered faster but it also divides the image into more parts and overall time is rising.

I think that the branching is getting closer for every pixel in each quad and also the GPU is dividing processing units more efficiently - it has to compute only 4096 quads this way - for the square of 1024 it is 262k quads. But not sure exactly why it is faster.

All I know about Fragmentarium is that it is also a program for rendering fractals.
I am more into programming my own solutions - just for the fun of the discovering not into clicking checkboxes and moving track bars.
And yes, sometimes it means to discover well known solutions once again. That's why I am here. To find out whether is something really new or all have been done before ten times at least.

[Sabine]

Hi and welcome, gurroa:)


I'm always glad to see some implementation in the 'real world' and what you do sounds really great!

Btw, if you're a programmer, you might like fragmentarium even more than I because since I am a mere non-programmer mortal I am happily doomed to check and slide, but for a programmer that program has a lot to offer:)


Hope you have a pleasant stay on FF!

[tit_toinou]

Yes. I suppose.

Nowadays I am rendering those huge images in small frame sized only 128x128 pixels.
All of mine fractals hlsl has at least two loops evolved.
And rendering smaller parts give you a lot faster result - comparing to larger part.

I have done some tests and find out that 128x128 is the average size possible.
Smaller is rendered faster but it also divides the image into more parts and overall time is rising.

I think that the branching is getting closer for every pixel in each quad and also the GPU is dividing processing units more efficiently - it has to compute only 4096 quads this way - for the square of 1024 it is 262k quads. But not sure exactly why it is faster.

I didn't know that I'll have to benchmark this, I'll also have to take into account the images joining process of course.