Weird shape for modeling challenge is stumping me... any suggestions?

[rkinusa]

An old teacher challenged me to model this shape and have tried to model it for time & am just completely stumped.

The task is not to just extrude the face otherwise I’d just cheat a bit and import the SVG and go from there.

It’s supposed to have a completely rounded 3D shape like a pill apart from the sliced off flat section on the left.

My first thought was to start with a pill shape and shift one of the caps over maybe and subtract a rectangle from the left over geometry, but nothing really seems to work.

The asymmetric curve between the bottom and the top is really throwing me off…

I feel mad stupid because it’s probably a really simple solution that I’m not seeing, but I just can’t get it out of my head now…

Any suggestions?

[DNorman]

Ha ha, old teachers are fiendish beings!
That is a tricky one because the shape could be interpreted in different ways, especially the transition between round and that flat side. Here is what I came up with (front, sides and top):

Edit: I made it a bit shorter like the reference.

[rkinusa]

Ahhhhhh I see what you’re going for there that’s definitely a solid method! I like what you’re doing with handling that transition… it looks like basically taking a quad sphere as a base and going from there.

I’ll certainly give it a go! Huge thanks!

Apparently there’s a way to tackle this with less than 10 “moves”… I’ve been messing around with modifiers with curves using an SVG input to see if there’s a procedural route.

But nothing can beat some good old box modeling haha. Thanks again!

[DNorman]

@rkinusa

it looks like basically taking a quad sphere as a base and going from there.

Yea that is what I did. There are different 3d forms that could have that profile.

@rkinusa

Apparently there’s a way to tackle this with less than 10 “moves”… I’ve been messing around with modifiers with curves using an SVG input to see if there’s a procedural route.

You got me there I did more than 10 moves - that is even more fiendish!

[md2000 (M D)]

I took the intersection set between a cylinder and a torus with zero center hole. Mate with a hemisphere.
However, it does have sharp edges.

The only way I can think of to do something with constant circular cross section is to use a circle quarter section as a background reference, and then subdivide and scale a cylinder cross sections progressively to match the reference. Scale each ring, move to align vertical edge, repeat… Sudivide for more detail.

This will be a tedious and inexact process, but each cross section will be circular. Don’t merge the top to a single point until the very end, so you can subdivide the extreme top as needed with CTL-R.

Here with the vertical edge also highlighted:

Maybe there’s a simple way to warp a cylinder with a profile to do this, but I don’t know Blender that well.

[md2000 (M D)]

Here’s Ortho and perspective views of the final product…

You can’t do Subdivision Surface modifier because it will smooth out the point.

[md2000 (M D)]

Fiddling with the new geometry nodes, I get this:

Which looks like this:

So basically, doing with GeometryNodes what I did by hand in previous entry.
(The starting geometry for the nodes is a cube)
After reflection about why(??) - “offset by Z” in the combine vector, basically adds z again to the original z making the top twice the height of the bottom hemisphere.

More Notes:
This does produce two unconnected meshes, top and bottom.
I apply the Geometry Nodes, then select the two edges and Mege By Distance to make it one.

This is the same setup but starting with an icosphere…

[Romanji]

Edit: …wait its supposed to be a 3D object?
Sorry, I was not paying enough attention.