Can you make a quick tutorial on the workflow? I installed it and have no idea how to use it lol
Yes sorry, this is the downside of releasing before documentation, I kind of have a specific workflow for it to be used.
I plan to add higher degree surfaces and all, but my first goal is to have a real workflow first. The focus on bicubic patches is very deliberate. You can basically construct any surface with several connected bicubic patches. Orders rise up quickly with the trims (work in progress), if you trim diagonaly a bicubic patch, you end up with order 6 bezier, so this is already hard to manage. Also the goal is to avoid to have some approximations done with some kind of solvers, but more keeping the topology good. All the feedback will tell if it is the way to go
I mean, It was supposed to happen in December along with a decent release, but I will do some quick screenshots, as the interest seems high
No worries. Two suggestions:
- Donât have different surfaces for creation, have just one. Then, in the geometry node input, allow the user to choose U and V degree on the fly. I did that all the time in Alias and itâs quite expected to be able to do it in any Nurbs software.
- Donât mix terms. Even if youâre in Blender, use actual Nurbs terms. Iâm not sure what you mean by âtrimâ increasing complexity, but in Nurbs software that doesnât happen (unless you use explicit tools for it) and thatâs the power of Nurbs. The trim edge however does increase in complexity, so maybe thatâs what you meant? EDIT: Yeah, reading again, thatâs what you meant! And I realize this is going to get difficult because an edge in Blender is not the same as an edge in a Nurbs softwareâŠ
But yeah, weâre probably talking across each other here, so Iâll wait for some videos or documentation to show up.
Apparently most of the assets are not in the addon right now and really donât know why. I should have called this version Alpha This is the first time I release something like that so the sketchiness is present haha
EDIT : This is fixed but not the only bug saddly
Keep going, this will be a game changer if you can get it working for the masses!
Blender recently had a big donation from Volkswagen. If you made them aware of this add-on you might get some dev fundingâŠ
Quick tutorial
- SurfacePsycho works by adding modifiers to mesh grids
- The modifiers are stored as assets, so add the path with the addon preferences
You should see this :
- The modifier stack is decomposed in 3 parts :
1 : The control grid deformers
2 : The Patch meshing (converts the grid to the patch)
3 : Others (mirror and combs for the moment)
- The most useful deformers are already in when you add a patch with shift+A
They are Continuities, to connect patches, and Trim 4 sides restrict or extend the patch parallel (in UV space) to its sides without any deformations.
- You can duplicate and apply them as you wish to shape your patch
- You can add other deformers from the asset browser or from the default modifiers (eg. Displace modifier to offset a patch)
More Tricks
Many tricks can be found and used with existing Blender tools
For example :
- Patch sculpting (classic sculpt mode)
- Some addons are very handy such as LoopTools and EdgeFlow (To straiten a row of control points for example)
- Snapping to points/edge with the control grid enabled
- I also use these x-ray settings to see the grid when editing
And finally you have this menu :
- The quick export saves the step file where the blend file is and with an automatic name. The classic way to export will be added later
- The surface probe is an object that you bind to some surfaces and it samples their curvature in every direction at the point where you place it. It plots this kind of clover shape, Also be careful it is sensitive to the patch resolution (and a sample radius displayed as a circle).
Saddly no, same as Mac OS, it is simply a matter of getting a build of Python OCCT for the os
So, I did the tutorial first so you understand a my the design. This is somehow different to other Nurbs editors partly because of technical constraints and partly on purpose. It kind of made more and more sens as I unrolled the idea and it is actually what motivated my work. Making a cheap copy of existing tools would never motivate me.
I cannot explain all the choices here but their is one core choice I can detail here : You may know that Blender kind of abandoned itâs Nurbs system and pretty much all attempts at reviving them failed. And for me their is a reason to that : 3D industry is kind of heading out of Nurbs. And the whole issue is this situation where there is a gap between mechanical industry who is like âPolygons ??? I donât trust that, I want .step !â. And almost all the 3D field is implicitly saying âNurbs ? To complex math, now with GPU we can use billions of polygons we donât careâ.
The amazing thing with using geometry nodes then is that it is simpler to learn than C, faster than Python, it is maintained by a trusted company, even better you can still use my system without the addon at all ! This provides a must have guaranty if a company want to use my system. They canât rely just on me.
So to answer :
-
That might be how it evolves, but in order to compute fast, right now I manipulate CP coordinates in parallel which in geometry nodes means connecting nodes for each CP. Maybe the parallel loop zone could solve that, but it is not announced for soon. Some algorithms might generalize to higher degree easily tho.
-
Yeah, I really want to have trims with curves on surfaces to some extent. I already have interesting results with some prototypes.
Any news about this addon?
Iâm still working on it everyday!
This week I did the initial support for planar surfaces, the week before I struggled with making it cross platform.This is definitely the hardest part yet, but it should be done soon.
Auto snap is working ! It selects the two closest sides to connect them, you no longer need to manually set them !
In special cases where the sides to connect are not the closest, you can still set them manually. Also it has his amount of glitchiness but this is expected since the closest sides can change quickly when moved and it depends on the order of the modifiers.
One of the many tests I am doing. This helps figure out many things. It is really not ideal for now, I need to add tools for triangular patches. Also it would be great to realize instances when exporting, because typically, two of blades are instances of the original blade collection which means it would exports only one of them.
C-infinity continuity
I just made Bezier curves with automatic order depending on the point count ! Here is the tool that allows to raise this order without changing the shape :
The equivalent for surfaces doesnât look impossible, however this is already quite slow when raising to high orders/resolution with curves (since it uses a serial loop on an attribute of size (order+1)*(resolution+1)), and it can only be worse for surfaces. I will think about it, curves were a good way of testing the technique
Just now seeing this. Very nice looking and definitely useful.
Iâve been researching ways to close the gap on CAD and polygonal modeling myself. Iâm modeling older style parts which were probably clay sculpted for sand casting. Translating that to CAD has its own challenges.
Version 1.0 will definitely not be out in December as I planned.
There are two many essential features with bugs or unpolished design. Also I am still working on making it cross platform and this is hard. I really want to promote a bit the first release, and make it a happy event so I will not rush it this week. I will just upload a v0.2 and only for windows
I just quadrupled the speed of patch meshing (~ From 4ms to 1ms on a 40*40 grid)
This was not a priority but it was the consequence of going one step closer to Patches of any orders.
Playing with the curves, I kind of decided cubic curves/surfaces will be separated entities from any-order curves/surfaces. This might be confusing for the user at first, but it allows to keep all the features I made for cubic entities, while transferring then slowly to the any-order entities (if possible). Also as far as I know the any-order entities will be significantly slower due to serial loops, so better to keep both for now.
Trim range and curvature combs are now available for any order curves !
Geometry nodes power exceeds all prophecies, I feel unstoppable