Hello, I don’t know if this is even possible, or if its been done before anywhere…
But my thought here, is if we could set a line on our paper, ( with nodes, ) that we could create arcs, and perhaps a false horizon with, where our gradient could radiate ourwards from.
I’m not a programmer, it in my mind sounds like a very tough task to do, but I think it could be a useful tool. 
oh interesting, I will give this a try
I have a hard time understanding the description of what you want. 
Can you elaborate and maybe make a mockup or paste examples to show what you have in mind?
Describe also your use case how you would use a tool like that.
Is it some sort of extension for Gradient Tool, or it has to work differently?
Why, you will need it as a tool and not just layer setup like @AhabGreybeard had made?
The better you describe your proposals with visual examples, the higher chance for high votes and developer taking interest in it. 
@Hoppa_Joel, are you thinking on something like Illustrator’s free form gradients?:
@Hoppa_Joel Like @Deif_Lou suggested you can use the mesh gradient or in the vector route you can also draw an arch with thick outline and then add a radial gradient to the stroke
Wow this is cool!
It looks like more advanced version of S-curves: S-Curves — Blender Manual
I wanted to make S-curve editing for Krita to allow drawing worklflow similar to the one used in Klaus movie.
My only concern is how to make it work with boolean operations. Well it would probably work by doing boolean operation on each data that is used to create it so probably doable.
Well, thats on my todo list as well, hehe. That thing in that klaus editor is similar to the masking tool that is present on different compositing apps. Basically you have a closed cubic bezier path and then you have another one that is offset outwards. The constraints on that offset path differ from app to app. But the thing is that they select opaque inside the main shape and then make a gradient from opaque to transparent, from the main shape to the outer shape. That’s the most tricky thing, making the gradient, and the approach would depend on how you constraint the offset shape shape I guess. One possibility is using diffusion curves, but they are expensive to compute.
Yes this should be kind of easy (relative to the rasterization step), either if you combine the vector shapes or you combine the results of the rasterization.
But better continue this on the irc if you are interested to avoid poluting the thread.
oh thank you for the replies, this sounds very cool
