Comment by socalgal2 | Hacker News Reader

Comment by socalgal2 | Hacker News Reader

Parent Full thread

Maybe the problem is just impossible or maybe AI assistance will solve it but it's crazy to me how complex 3d software like Blender/Maya/3DSMax/Houdini etc still are. There are 1000s and 1000s and 1000s of settings, Deep hierarchies of complexity. And mostly to build things that people built without computers and only a couple a few tools in the past. I had hoped VR (or AR) might some how magically make this all more approachable but no, it's not much easier in VR/AR. The only tool I've seen in the last 30 years that was semi easy was the Spore Creature creator though of course it had super limits.

I guess my hope now is that rather than select all the individual tools I just want AI to help me. At one level it might be it trying to recognize jestures as shapes. You draw a near circle, it makes it a perfect circle. Lots of apps have tried this. unfortunately they are so finicky and then you still need options to turn it off when you actually don't want a perfect circle. Add more features like that and you're back to an impenetrable app. That said, maybe if I could speak it. I draw a circle like sketch and say "make it a circle"?

But it's not just that, trying to model almost anything just takes soooooooooo long. You've got learn face selection, vertex selection, edge selection, extrusion options, chamfer options, bevel options, mirroring options, and on and on, and that's just the geometry for geometric based things (furniture, vehicles, buildings, appliances). Then you have to setup UVs, etc.....

And it gets worse for characters and organic things.

The process hasn't changed significantly since the mid-90s AFAICT. I leared 3ds in 95. I learned Maya in 2000. They're still basically the same apps 25-30 years later. And Blender fits right in in being just as giant and complex. Certain things have changed, sculpting like z-brush. Node geometry like Houdini. And lots of generators for buildings, furniture, plants, trees. But the basics are still the same, still tedious, still need 1000s of options.

It's begging for disruption to something easier.

Couldn't this be said for any creative software, e.g., After Effects, IDEs, DAWs, NLEs, Photoshop, Illustrator, etc... and even a few non-creative applications like Word and Excel? (To your point, I do think 3D modeling software is the most complicated of all of these, but I think as a general rule, the complexity is more similar between these software categories versus other categories like Mail, Notes, etc...)

For my part, I think this is because to do good creative work you need minute control, and minute control essentially just means adding lots of controls to software. Sure you can mediate this with AI, the learning curve especially, but that only really helps unskilled workers and hinders skilled workers (e.g., a slider is a more efficient way for a skilled user to get an exact exposure, rather than trying to communicate that with an AI). And I don't really think there's a need for software where unskilled users have a high-degree of control (i.e., skill and control are practically synonyms).

I see your point. Yes, all those types of software are complex. And, maybe that's just the way it has to be. But, take the non-software versions. Sculpting a bust out of clay is certainly a difficult skill, but it doesn't take 3000+ hierarchical options. All it takes is a pile of clay and 2 hands. Maybe add a 1 to 5 tools. Similarly, building lots of wood furniture is a skill but doesn't take 3000+ hierarchical options. It takes just a few tools. Drawing a diagram on paper (Illustrator) takes a pencil, a ruler, maybe a compass.

I just feel like there's some version of these tools that can be 100x simpler. Maybe it will take a holodeck to get there and AI reading my mind so that I knows what I want to do without me having to dig through 7 levels of menus, sub sections, etc....

I think this point trends into HCI (e.g., https://worrydream.com/ABriefRantOnTheFutureOfInteractionDes...) but I think the overarching phenomena you're describing is that custom interfaces are usually superior to adaptions to the mouse/keyboard.

There are many areas where custom hardware interfaces are popular used in conjunction with software:

- MIDI controllers (note both for the piano keys, and for the knobs/sliders, and even foot pedals)

- Jog wheels for NLEs

- Time-coded vinyl https://en.wikipedia.org/wiki/Vinyl_emulation

- WACOM Tablets

All of these custom hardware interfaces accomplish the same thing: They make using the software more tactile and less abstract. Meaning you replace abstract-symbol lookup (e.g., remembering a shortcut or menu item) with muscle memory (e.g., playing a chord on a piano).

So TLDR, the reason that we don't have what you're looking for is that we don't have a good way to simulate clay and wood as hardware that interfaces nicely with software.

Note there's a larger point here, which I think is more what you were getting at. I think people sometimes expect (and I expected this when I was younger), that computers could invent new better interfaces to tasks (e.g., freed from the confinements of physics). Now I think this is totally the opposite, that the interfaces are usually better from the physical world (which makes sense if you think about it, often what were talking about are things that human beings have refined over thousands of years), and that enforcing the laws of physics usually actually makes things easier (e.g., we've been dealing with them since the moment we were born, we have a lot of practice).

Finally also note that custom hardware interfaces only tend to help across one axes (e.g., a MIDI controller only helps enter notes/control data). The software still ends up being complex because people also want all the things computers are good at that real world materials aren't, like redo/undo, combining back together things that have been broken apart, zooming in/out, seeing the same thing from several perspectives at once, etc...

PS I don't even know if the Holodeck or mind-link up would really help here, it's possible, but it's also possible it just difficult for our brains to describe what we want in a lot of cases. E.g., take just adjusting the exposure, you can turn it down, oh but wait I lost the violet highlight that I liked, how can I light this scene and keep that highlight and make it look natural. I don't know maybe this stuff does map to Holodeck/mind-link, but it's also possible that just having tons of options for every light really is the best solution to that.

> It's begging for disruption to something easier.

This assumes the complexity is incidental rather than inherent. I think the problem is similar to how Reality has a surprising amount of detail[1]. AI will likely eventually make a dent in this, but I think as an artist you tend to want a lot of granular control over the final result, and 3D modeling+texturing+animation+rendering (which is still only a subset of what Blender does) really does have a whole lot of details you can, and want to, control.

[1] http://johnsalvatier.org/blog/2017/reality-has-a-surprising-...

Reply on news.ycombinator.com