Thanks guys! It's the combined efforts of many people exploring the algorithms within these forums that make this field so much fun
The video has had a crazy number of views over the past couple of days so maybe we'll see a few new faces exploring these forums too.
Glad to see this - hope he releases the code. Tom's Mandelbulb shader got me started down the fractal rabbit hole. Fun ride!
It's great to hear that! I still have a few more things I want to achieve with it first, but then I will open-source it. No timescales yet though.
and what exactly is he doing with the primitives ?!?!?!
It's a geometry trap - after each iteration pass the iterating point does a distance estimation to a primitive shape and the minimum distance stored to get the geometric union effect of the structure. Then by being able to rotate and scale the primitive shape you can get a huge variety of new structures from the same base fractal algorithm.
Compared to something like Mandelbulb3D there are only really a handful of different algorithms, but each have been parametrised and augmented with rotation, scaling and additional transformation features that there is still a massive parameter space to explore (as seen with some of the more crazy looking Kleinian structures in the screen captures).
In reply to your personal message ck about the hybrids - I did actually explore that a few years ago with my PixelBender scripts. Both as alternating between two formulas between iterations, and running both in parallel and then blending their distance estimates. However, I found that more manageable results could be found by augmenting the single fractal formula.
An important factor in developing FL was that the user interface had to be very clean and enjoyable to use (I get very frustrated if the UI gets in the way when I'm trying to explore something creatively
Everything has to be immediately responsive with very accurate control because otherwise many interesting structures could be missed since a 0.001 change on one parameter could be when everything suddenly pops out.
Also at the time the GPUs I had access too would lock up if the shaders became too complex, so I've not implemented the full Hybrid features. Instead I've focused more on trying to build the best rendering engine I can. Even with the fairly standard base fractal types there is still a vast parameter space to explore creatively!
A more dynamic node based render pipeline would certainly be one approach to properly implementing hybrids - the output of which compiles to a single shader and building the UI accordingly. Not something FL can do in its current form just now.
It's all good fun though