Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

A wonderful app that, I find, benefits from LLMs to help you figure out how to use it. That or you more or less have to dive in for months and months. The people on YouTube that make Blender look so easy … Blender is really all they do. ;-)

At some point I see an LLM more or less integrated into the UI.

At some point I see whole apps written so complexly that an LLM is the required interface.



> That or you more or less have to dive in for months and months.

Waaaaay back in the Quake 3 days I had a pirated copy of 3D Studio Max and decided to try and make a Quake (or maybe it was Half-Life) character model. Found an on-line tutorial that step by step showed you how to setup Max with a background image that you "traced" over. So I grabbed the images of front, back and side views of a human, loaded them into the various view ports, then drew a rectangle from which you extrude a head, arms and legs. Then you manipulate the block headed human mesh to fit the human background image - extruding more detail from the mesh as you go along. In one day I had a VERY crude model of a person. I also found out I dont like 3D modelling. Though I'd say a person who really enjoys it would pick it up in a week or two with good tutorials.

LLM's just cut out the learning part leaving you helpless to create anything beyond prompts. I am still very mixed on LLM's but due to greed and profit driven momentum, they will eat the world.


One problem I find is that a lot educational content has moved into YouTube and videos (monetization be damned). I have no time to watch 10mins of rambling and ads for a quick tip, LLMs are great at distilling the info. Otherwise, I agree, deep knowledge building only happens through doing stuff…


Oh boy. Giving Blender instructions in English, on every level of detail, could be quite a workflow. Like, have it generate a range of chairs, pick the mesh you like, then start asking it to tweak parts of it until you are satisfied. Add finishing touches manually.


In general, model re-topology almost always requires a manual editing process post sculpting and texturing. There have been countless attempts to automate the process, but it usually requires working through an obscure book about rigging with secondary transforms etc.

Could check out https://extensions.blender.org/add-ons/mpfb/ for low-poly character mesh generation tools. =3


Why is this being down voted?

If you are pushing vertices around to make a model, you eventually have to think of topology issues if you ever want to make a good animation.

And then when you have a good topology... Then You need to rework the UV Maps, Normals and other such data so that it's all consistent.

LLMs are not the right tool for this. I'm sure some AI retopology tool can come around but there is a HUGE difference between a model designed for stills, a model designed for animation and finally, a model designed for close up facial animations.

The topology differences alone are insane.

---------

Now maybe in the future an Ubermesh (like the Uber shader / principled shader) can be made to make these processes consistent. And then tooling can be made over the hypothetical Ubermesh. But today it's a lot of manual retopology that requires deep thinking about animation details.

Ex: an anime style mouth doesn't usually move the jawline (Think Sonics mouth: its basically a moving hole that jumps around his face). Meanwhile, western 3D animation (Baldurs Gate and the like) obviously have fully made lip sync including phhfff, eeee and ahhh sounds.

The 'Kind' of animation you want leads to different mesh topologies. It's just the nature of this work.

And people get ANGRY when you ex: animate Sonic with traditional teeth and jawline (see 'Ugly Sonic'). You can't just willy nilly pick one style randomly. You need purposeful decisions.

--------

I guess if you are a pure sculptor / stills then you don't have to worry about this. But many 3d projects are animated, if not most of them. The retopology steps are well known.

LowPoly look will almost certainly require manual effort (these topology problems tend to disappear with more triangles).


Blender is often used to project the high-granularity sculpted detail onto a low-poly (rigged and clean topo) Normal map. i.e. you can get great looking assets that won't hit your poly budget as hard. When it works its great...

People have feelings, and sometimes may misinterpret others intent. I am happy we've met if we both learned something. =3


Oh yeah, I was agreeing with you. Just providing context for others if they didn't understand what was going on.


Huh, TIL https://github.com/ahujasid/blender-mcp - do you use this?


I tried it and it's really not very good. It "works" but models can't seem to do anything complicated. Claude 4 Opus btw


Thanks for pointing that out.

I have not tried it. Instead I have been asking Claude (etc.) "How do I create a repeating triangular truss..." or what-have-you. And then I follow the steps they list.




Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: