I think i have to learn how to use tensor fields…
Anyone knows a good place to start?
Here’s my fun with circle transform and pr_deform on top:
It’s a bit buggy at boundaries but could be fun to have this kind of things added to the script. Except it would be heavy parameter (GUI or CLI) wise if i want to give more control to the user.
Any advice? Which road would you take? I haven’t explored all the “deformations” folder yet.
Looking good so far. On my mobile phone, I do not notice the boundary issues, but on the laptop or larger screen, yes, they are there at the image edges, as well as the flakiness on the clouds’ edges.
In terms of parameters, I would recommending having a light and heavy version, or you can play with the GUI to organize them as @Reptorian has in the past. Takes a little while to understand though.
That’s a doozy, but I recommend you that use pen and paper, draw out the logics, debug always in your head. That’s how I manage to work with 1000+ variables.
I’ve just thrown a copied circle transform line from the GUI in the middle of the script, just to test it, and this was the result, so nothing really serious. The flakiness/fuzziness is intentional though, but optional. Doesn’t look that good without it. It’s also a bit more painterly (chalk/pastels maybe?), which i like. There’s probably a million ways to make these clouds anyway.
Yeah well i remember the GUI status update nightmare.
I also saw a few filters which could show/hide options. I wonder how you handle the vars for hidden options. Probably the same as if they were shown, and leave them as default if not touched.
I’m also tempted to make a version which would use a binary image as shape, but i’m not sure i can succeed.
I wonder how that would look like? I sure don’t want 1000+ vars in there, even if hidden… But i’d like to throw some more shaping options though, so that probably means more sliders if using deformation filters. Or i could try using a guide to place objects around (like, a curve) ? don’t really know how to do that, so if someone know a good example…
I feel like stepping into unknown territory each time i start something…
In my case, they’re unused when it’s not needed or they’re used to preserve previous states or update states. Previous states can be used to check if a new image is to be created or use from $_persistent.
Is there um, a way to make this shorter?
if same('$1','z') m __$0_split:"+crop[0] 0,0,$""1,0,0,100% crop[0] 0,0,0,0,0,{$""1-1}"
elif same('$1','y')' m __$0_split:"+crop[0] 0,$""1,0,100% crop[0] 0,0,0,{$""1-1}"
else m __$0_split:"+crop[0] $""1,100% crop[0] 0,{$""1-1}"
fi
split command doesn’t have a way to split exactly at nth point. Just equal sized split or every nth point.
Have you tried with columns (or rows)?
columns:
x0[%],_x1[%],_boundary_conditions
Keep only specified columns of selected images.
'boundary_conditions' can be { 0:dirichlet | 1:neumann | 2:periodic | 3:mirror }.
Default values: 'x1=x0' and 'boundary_conditions=0'.
Example:
[#1] image.jpg columns -25%,50%
No. But, I think I found a idea with what you gave.
Alright, don’t know how i managed to write this crap, but i did it somehow…
TBH there’s probably a lot of useless or meaningless stuff in there.
Got so much weird but fun stuff doing this too.
ellipse_tree:
# ... or at least some kinda trunk-ish thing
1000,1000,1,3,255
X=50
Y=100
N=1000
H=.95
W=1
r=6
rip:=$r*.8
#~ O=.01
O=.1
#~ O=1
undul=.1
a=.1
repeat $N {
ellipse[0] \
{$X+(u(20,25)*cos($>/w))*sin($>/u(95,100))*$undul}%,\ #x
{($Y-($>*.1)*$H)}%,\ #y
{$W+lerp($>/$N,$</$N,1)}%,\ #R
{$r*sin($</h)}%,\ #r
{($a*$<)*sin($</u($rip))},\ #a
$O,\ #o
0xFFFFFFFF,\
170,100,100
#~ w
#~ wait 50
}
d
Anyway, now i wonder:
Some weird stuff:
If you want to make branches, you can have a loop that uses x,y coordinates of a random ellipse. And you have iterative approach or recursive approach. Recursive approach means to call itself.
You can try blurring at local parts. That’s hard to do.
You can make another layer of mathematical calculation to create bends.
Also, it seems that you have %. That means you definitely need to use multiple math parser call. Another option, it to create variables and apply the math in a single math parser and use those variables next to %.
Yes, that’s what i often do, it’s just that at the beginning of a new … script(?) i don’t know yet what i can use as vars, so i just put everything in the command.
Do you mean the command blur? Or more maths to just screw things up here and there?
Yes, snapshoting x,y (and the rest) is what i was going for ( and that’s why i use ellipses instead of a spline, so i can find x and y). But i’m not sure i can draw this thing going left or right. Not yet, at least.
Some color variation:
Maybe i should eat more L-Systems for breakfast.
Yes, you could do it like this.
x_coords,y_coords=${1-2}
repeat $number_of_branches {
find_new=1
repeat $loops_per_branch {
if $find_new&&$criteria
nx,ny:= # Calculate here
find_new=0
fi
}
x_coords,y_coords=$nx,$ny
}
More maths.
mmm I think i’m going to restart from scratch with just the branching. There’s too many things going on with these cos & sin it’s unreadable. And if i can’t do the branching, then there’s no point in continuing anyway…
I rewritten codes as a process. For example, my bin2dec. I rewritten it like 5+ times, and I still know where to improve on.
In my case i have big trouble with maths so i can only check the visual results.
That means i can’t spot errors or superfluous calculations.
I hate going out of the canvas for example, but that’s probably not the worst.
You can use print() on small images to at least see what’s going on. Or modify variables to test your assumptions. display works too.
Yes. When i use repeat (i do a lot), i use w to see the steps and spot unwanted behaviour. But i can’t see anything if things go out of bounds, so maybe i’ll go with print or echo vars in the terminal.
Here’s what i got so far following your example:
You could do a 3d version of this with metaballs. Imagine making 3D trees in G’MIC.
