Image not available

843x727

blender vs houdini.png

๐Ÿงต Untitled Thread

Anonymous No. 951918

How do Blender's simulation geometry nodes hold up against Houdini? Obviously Blender's way behind the curve, but does it seem like it might catch up in the future?

Pls try not to shit-fling, talk about actual specifics and features.

Anonymous No. 951921

The devs promised to adapt all the object modifiers to to nodes. The "everything nodes concept".
When that's done it could hold it's water.
Right now I doubt anyone will see it as competition to Houdini.
It doesn't have to be though, and it's not the same focus.
While many tutorials show you how to create something from scratch in geometry nodes that's not the best way to use it which is about creating your own tools that modify a mockup until becomes a final model.
Say do simple box modelling of a floorplan and turn that into a house. Distibute several of those boxes onto a plane and turn that into a city.

Anonymous No. 951922

>>951921
>Say do simple box modelling of a floorplan and turn that into a house. Distibute several of those boxes onto a plane and turn that into a city.
Yeah, and with simulation nodes you could even do something like auto-distributing pre-made pieces step by step using a wave collapse approach.

Image not available

760x725

Screenshot_202307....png

Anonymous No. 951947

>>951922
>wave collapse
Can you explain/summarize this in english? I hear it enough that it's probably a common phenomena I would recognize. but don't know the name for.

Anonymous No. 951952

Doubt it'll 'catch up' to Houdini in all departments esp FX. SideFX are decent Devs and updates to Houdini are usually substantial and meaningful. I can see it being an interesting alternative for basic and medium complexity procedural modelling set ups. Beyond that, they need an answer for VEX or as a stopgap a node that executes python on the geometry - nodes are cool but they're very tedious in some scenarios especially as complexity ramps up.
Houdini indie is also priced really well.

Anonymous No. 951954

>>951947
It's an unfortunate name. But the name and the confusion with it is part of the reason it's popular.
First off idk if it can be done with geonodes.
AFIK it goes like this
>define rules of which sort of tile can go next to which other and where.
>forest > grass > sand > river
>river > river > lake
>path > path > bridge, etc....
>place random tile on the map
>surrounding tiles are filled with a mask what they can be
>those surrounding those mask tiles are then filled with another mask of what the can be
>repeat until the mask opens up all possibilities again
>place another random tile and take into account the mask
>repeat until all tiles are defined.

What's great about it is you can create your own map by hand and leave some tiles undefined and it will generate variations.
You can even roll back the algorithm if you save the intermediate results and get variations that are very similar.

Anonymous No. 951959

>>951954
Yeah, it's basically this:
>Every piece on every one of its edges it has a "list of allowed neighbor pieces"
>you start in a random place with a random piece
>you choose the spot on the map that has the "least amount of choices" and place a randomly chosen (allowed) piece in that spot
>the more pieces you place next to each other the more you'll narrow down their neighbor choices
>repeat until map is filled

Anonymous No. 951964

>>951947
it literally doesn't mean anything
it's a bunch of gobbledegook nerds spew to make themselves look smart

Anonymous No. 951966

>>951952
A minimalist C++ node with a straightforward API would blow VEX out of the water, you try to do anything custom in VEX that doesn't align with Houdini's pre-existing data structures (like building or recursing down a tree) and the performance is worse than python.

Anonymous No. 951969

>>951966
You do not understand the joys of VEX

Anonymous No. 951973

https://www.youtube.com/watch?v=2SuvO4Gi7uY
Superpositions, Sudoku, the Wave Function Collapse algorithm.

Anonymous No. 952279

I don't think it ever will unless SideFX make really bad decisions in the future but the nodes in Houdini and the built-in VEX functions are way too advanced compared to what you can currently do in nodes/Blender.
Just the system and data flow of Houdini makes stuff possible that you just can't do in any other software.

Anonymous No. 952290

>>951918
blender has ALWAYS copied features in a half-assed way. it will not be different this time.

Anonymous No. 952394

>>952290
I know first hand that sidefx does the same thing

Anonymous No. 952480

>>952394
post example

Anonymous No. 952481

>>951966
>minimalist C++ node
wouldn't this require blender to ship with a built-in C++ compiler? Could a python node suffice?

>>951952
>>951969
>>952279
What makes VEX so good? I've never had anything to do with it, all I know is python, C and C++

Anonymous No. 952487

>>952481
>What makes VEX so good?
It's built around the manipulation of data for point and primitives. It's a little programing langage for the manipulation of whatever you want (points, edges, primitives, normals, etc...). You can give them attributes and group them. Add to that all the built-in functions to operate all that and you have something powerful.

Anonymous No. 952497

>>951918
maybe if blender hires more PhDs

Image not available

277x122

capcha.png

Anonymous No. 952504

>>951973
Thanks for sharing that.
>pic unrelated
https://www.youtube.com/watch?v=KPoeNZZ6H4s

Anonymous No. 952523

>>951966
this would require blender to build a c++ api (which they really need to do at some point), but that's a multi-year undertaking. a python node now would be a quick, pragmatic, fix.

>>952481
VEX is a C-style scripting language. it's parallelised by default on the cpu (across points, vertices, or primitives) and requires no real boiler plate since it is somewhat limited and well integrated into hou. you drop a node in-line and just type what you want to do.
it won't let you define custom data structures like a general purpose programming language would but most things you need should already be covered (non-square matrices is a notable omission)
if you already know C you should be able to pick up in a day or two. it takes a second to adjust to the fact that, by default, you're working inside a for loop running across the geometry. you can run serial if you want.
you also have openCL and python in hou if you need them.