KittyCAD - next generation CAD for those who like both code-based and UI-based design

Hey folks,

I guess there are not that many here who prefer a code-based approach to design, but for those who might have some interest, there is a new CAD tool project called KittyCAD.

It is a similar concept to OpenSCAD in some sense, but the method of building an object can be done like Fusion 360, Onshape, etc. The big difference is that while you are designing your components, there is a code-based representation of that object that is automatically built at the same time. And better yet, you can move back and forth between the UI view and the code view to change the model, using whichever technique seems the most intuitive, precise, etc. for that operation.

Unlike OpenSCAD, which is based on CSG, KittyCAD is based on B-rep (Boundary Representation), which is, it is said, far more flexible than CSG. There’s a wikipedia article on B-rep: Boundary representation - Wikipedia

Anywho, you can try out a web browser-based version of it here:

The main web site is kittycad.io. You can get the alpha version of the CAD client here:
KittyCAD: Download KittyCAD Modeling App

There are clients for both Windows and MacOS. There is a Linux build as well, but it currently is broken, and it may be 6 or more months before that problem is resolved due to some compatibility issues which are out of their control.

Anyway, I’m really jazzed about this, and think it has the potential to be better for me than all of the current options out there.

1 Like

Thanks for posting!

Cheers

1 Like

Very interesting. I have just started looking at it, and it may be a bit too early to make a video about it, but it is an interesting concept.

Their challenge, like that of most software companies, will be commercialization.

1 Like

Yeah, it’s definitely too early for a video. I give it a year before it’s really usable.

I found this roadmap doc from their #expectations channel:

They’ve made quite a lot of progress. Check out this video:

1 Like

I’ve been playing with it a bit recently, and have filed a few bugs on their github.

I can see a lot of possibilities here for this becoming my go-to, because it “fixes” a lot of the usability issues that I have with OpenSCAD. For example line (polygon) construction in KittyCAD is much easier to code, and better yet, you can label objects in the model, and use those labels to place other objects (instead of purely through explicit calculations as in OpenSCAD). I believe there will be built-in ways to measure as well.

But it’s still too creaky to really be usable for creating usable models, either through code, and even less so by point and click. Yet.

They are at version V0.29.1 now and are thinking a lot about their V1.0.0 release. It’s not really clear when that will be, but at the current rate, my guess is end of the year.

1 Like

The Zoo team has made quite a lot of progress in the AI part of this tool. If you follow the Discord link to this video, they demonstrate describing a shape in English, and using AI to create a parametric model from it, written in their new language (KCL), using good variable names, with comments describing what it’s doing. That code is subsequently editable. It’s very impressive!

1 Like

The zoo team released v1.0 today.

It has a lot of improvements since my last post.

  • The syntax of KCL is cleaned up and a lot easier to read now, but still takes a bit of getting used to.
  • Boolean operations are available now. They mostly work with a few caveats that will be fixed over time.
  • Modularity is much improved, so you can now import other kcl code modules, STL, STEP, 3MF models, and do some renaming for readability.
  • The AI aspect is easier to use now, but personally I think it’s still not ready for general use. It feels very limited, unless you have it design something very well known in its training data, e.g. “design a gear with 25 teeth, 30 mm in diameter with a 5 mm shaft hole”
  • The point/click interface is partially there for most operations, including sketching where it’s quite good now.
  • Filleting by code and point/click is available, but it has some pretty severe limitations at the moment, and is also buggy/unstable. There are some outstanding issues that are being worked on to resolve the instability first. The removal of limitations of the fillets themselves will come later.

I have been playing with it for the past year, with many small models. My overall experience with it remains pretty much the same as last year - it has great potential to be my"one stop CAD", but the bugs and limitations in the functionality are keeping me from really adopting it completely.

It keeps getting better and more fleshed out over time. I like the KCL coding language more now than before, and I thought it was pretty good to begin with. It’s quite elegant!

The downside is that it seems the bugs I have reported don’t get fixed in a timely manner. Some fairly significant ones I posted months ago are still not addressed.

My hope is now that v1.0 is out that they will have time to address some of the functionality bugs that have slowed me down a lot in using it for real models I want to 3d print.

The app remains free to use, and will always be free for hobbiest-level users. There is a limitation on the the AI usage to 100 prompts/month, and they retain the right to train their AI on your models at the free tier. None if this is an issue for me because I will almost never use the AI, and I don’t mind them training on my models (which are very specialized, and probably useless to them!)

All of this comes with a significant caveat – if you write your models in KCL, and then Zoo fails as a company, your KCL code will have to be translated – somehow – to another CAD system to be usable, because most of the CAD brains are in the CAD engine which runs on their cloud. The client is mostly a (fancy) viewer of the engine’s rendering.

Conversion from KCL to something else won’t be easy. There is some discussion on their Discord that if the company does go down, they would likely open source their CAD engine so that customers wouldn’t be stuck, but it wasn’t a promise. From what I understand, they have plenty of funding at the moment, and there’s no current danger of going under.

1 Like