Biosuit WIP

[Mathaeus]

Hello,
just wanted to show my 'transition wip for fun', perhaps belonging to two forums here, anyway for now it's much more Houdini thing. So it's rigged and uv-ed model exported from Maya, Houdini part is clothing. So, except old geometry from XSI 4 or 5, everything is 'transitional'.
It's inspired by modern astronaut's 'biosuit', also I'm trying to add some erotic elements, bondage, stripes and such. Houdini part is of course completely procedural (because my direct modeling skills in H are somewhere around drawing a curve ). For now all these stripes are product of something close to 'deform by uv' in ICE, but with a bunch of additional methods like so-called 'convex surface smoothing' of target surface, to get wrapped curvature. Basically a plain repeat smooth, but projected back to target in each iteration in case the result interpenetrates with target, this gives a sort of smooth convex hull after many iterations.
Have plans for procedural created wrinkles on tight cloth, while have no exact idea how to do that .
Harness system is pretty much interactive, by moving small red spheres, entire things adapts to body.
In good old times I'll do all that in just one software . Anyway to stay on positive attitude, have only a nice words for H optimized evaluation, once complete circus of VOPs is evaluated, small changes are evaluated almost immediately. Also, Maya uv tools, like cut-sew workflow, selective unfolding and optimizing, allows to perform a very, very precise operations with UVs.
Hope I'll have a time to finish all that, into nice render.

[anhungxadieu]

cool! the rig look complicate

[Rork]

Thanks for keeping up the posts on your project, very cool to see the developments on this.

And I agree on your sentiments on Maya and Houdini. Maya has a great set of modeling and UV tools nowadays, there's even some great Bonustools for stuff like that.
And Houdini shines with all big data sets. It will just work, and render as well. Something Maya still seems to have issues with after all those years.....

rob

[Mathaeus]

Thank you guys

cool! the rig look complicate

A lot of of pure optional stuff is there, it's more a single-user concept to keep literally everything in just one structure, from guides, node made IK solvers, layered hierarchies, whatever else. Found that way easier to manage, for personal use. I'll always be able to detach or delete some component, while there is no need to do that for now. All you see is playing on 60 - 90 FPS, depending on what is displayed.

[rray]

>>Have plans for procedural created wrinkles on tight cloth, while have no exact idea how to do that

Sounds interesting, would probably require tension computation with some smoothing steps to get a nice tension direction value for the points. But how to go from there to actual folds... no idea either

Harness system gets my red dot award

[Bullit]

Does Houdini have compounds to share like Softimage ICE? The question popped up when Mathaeus wrote about deform by UV.

[Mathaeus]

Does Houdini have compounds to share like Softimage ICE? The question popped up when Mathaeus wrote about deform by UV.

Well as you mentioned 'sharing', I'd say yes and no.
Yes: There is an equivalent of ICE preset manager window, so you are able to drag drop a custom HDA (Houdini Digital Asset) into network.
No: Creation, managing and sharing, that's a way more complex than ICE or Fabric, which must be great if there is dedicated TD in facility, so much more options and all that. However it's a way harder to manage, much easier to do something wrong, for example, I've managed to mess-up few factory HDAs while ago, so had to re-install Houdini to fix. Parameters are linked by expressions, not always by plain connect, so, copy paste of some nodes from one HDA to another, results in zeroed parameters. Evaluation is not same when you're into HDA or out. Finally, installing HDA from Orbolt site, requires a bit of IT knowledge (at least it was few years ago).

As matter of fact, people on H sites prefers to share H scenes, much more than HDAs.

On positive side, copy-paste between scenes works flawlessly, for much wider range than ICE (as (almost) everything in H is 'networked'). Personally I'm trying to avoid HDAs in favor of many smaller, per object networks. For possible systems of compounds, I'd prefer Fabric, which here offers a real multi-platform, together with robust and simple, ICE style of managing.

[Mathaeus]

A few small tricks to share...

Transparency in viewport: for getting that old school alpha like helmet in first pic, still not doing anything with materials, it seems to be enough to create attribute called Alpha and set something between 0 and 1, somewhere in SOP network.

Convert VDB to polygons, still having nice normals: thing in pic is VDB converted to polygons, however the default VDB to poly converter does not playing nice when it comes to normals, it seems they are just interpolated around meshed result. So there's trick to supply normals directly from VDB. Comparison pic shows default and 'generated' normals. Volume gradient' in VOP actually is a sort of 'VDB normal', it could be used in similar way as mesh normal.

[mc_axe]

cool stuff

[Bullit]

Does Houdini have compounds to share like Softimage ICE? The question popped up when Mathaeus wrote about deform by UV.

Well as you mentioned 'sharing', I'd say yes and no.
Yes: There is an equivalent of ICE preset manager window, so you are able to drag drop a custom HDA (Houdini Digital Asset) into network.
No: Creation, managing and sharing, that's a way more complex than ICE or Fabric, which must be great if there is dedicated TD in facility, so much more options and all that. However it's a way harder to manage, much easier to do something wrong, for example, I've managed to mess-up few factory HDAs while ago, so had to re-install Houdini to fix. Parameters are linked by expressions, not always by plain connect, so, copy paste of some nodes from one HDA to another, results in zeroed parameters. Evaluation is not same when you're into HDA or out. Finally, installing HDA from Orbolt site, requires a bit of IT knowledge (at least it was few years ago).

As matter of fact, people on H sites prefers to share H scenes, much more than HDAs.

On positive side, copy-paste between scenes works flawlessly, for much wider range than ICE (as (almost) everything in H is 'networked'). Personally I'm trying to avoid HDAs in favor of many smaller, per object networks. For possible systems of compounds, I'd prefer Fabric, which here offers a real multi-platform, together with robust and simple, ICE style of managing.

Thanks Mathaeus. Hope they can improve that. It was amazing the ease of sharing effects in ICE.

[Bullit]

Mathaeus what is your opinion on use of Digital Assets as this guy?

You don’t need to just model stuff in Houdini. Your brain will definitely collapse, once you go over 100 nodes. That’s why I apply a different strategy here. To get the best results out of Houdini you have to use a Digital Assets mindset, thinking in a modular way. With this approach, you encapsulate the massive number of similar (or different) actions into one node, which helps to achieve the necessary result. This is the basic rule, which will help you to understand the incredible power of Houdini.

For example, when I model something I use a Digital Asset called 3 Step Extrude. The name pretty much sums up the function. Let’s say I have a cube, which requires me to compile 9 steps in extruding. The final result should be put into a group and two separate materials must be applied (one for the sides, one for the frontal polygons).

So to complete this task I require three nodes of 3 Step Extrude or 11 standard nodes like Poly Extrude, Groups, Materials. And that’s just the most obvious usage case. Imagine you’re working on a Hard Surface detail with 20 extrusion steps. Working in 3ds Max and similar programs you’re acting as a sculptor, looking at the draft and modifying it. You’re not worried about the number of actions you need to do to achieve the final result. With Houdini this approach just doesn’t work. People start to use Houdini when they want to achieve great results and cut down the production time. That’s why it’s so important to work on your own library of Digital Assets. My own library holds over 50 assets just for modeling alone. It’s a great help in my daily workflow.

https://80.lv/articles/building-indie-g ... dern-tech/

[SamHowell]

A few small tricks to share...

Transparency in viewport: for getting that old school alpha like helmet in first pic, still not doing anything with materials, it seems to be enough to create attribute called Alpha and set something between 0 and 1, somewhere in SOP network.

Convert VDB to polygons, still having nice normals: thing in pic is VDB converted to polygons, however the default VDB to poly converter does not playing nice when it comes to normals, it seems they are just interpolated around meshed result. So there's trick to supply normals directly from VDB. Comparison pic shows default and 'generated' normals. Volume gradient' in VOP actually is a sort of 'VDB normal', it could be used in similar way as mesh normal.

Thank you for this. A very useful tip.

[Mathaeus]

Mathaeus what is your opinion on use of Digital Assets as this guy?

You don’t need to just model stuff in Houdini. Your brain will definitely collapse, once you go over 100 nodes. That’s why I apply a different strategy here. To get the best results out of Houdini you have to use a Digital Assets mindset, thinking in a modular way. With this approach, you encapsulate the massive number of similar (or different) actions into one node, which helps to achieve the necessary result. This is the basic rule, which will help you to understand the incredible power of Houdini.

For example, when I model something I use a Digital Asset called 3 Step Extrude. The name pretty much sums up the function. Let’s say I have a cube, which requires me to compile 9 steps in extruding. The final result should be put into a group and two separate materials must be applied (one for the sides, one for the frontal polygons).

So to complete this task I require three nodes of 3 Step Extrude or 11 standard nodes like Poly Extrude, Groups, Materials. And that’s just the most obvious usage case. Imagine you’re working on a Hard Surface detail with 20 extrusion steps. Working in 3ds Max and similar programs you’re acting as a sculptor, looking at the draft and modifying it. You’re not worried about the number of actions you need to do to achieve the final result. With Houdini this approach just doesn’t work. People start to use Houdini when they want to achieve great results and cut down the production time. That’s why it’s so important to work on your own library of Digital Assets. My own library holds over 50 assets just for modeling alone. It’s a great help in my daily workflow.

https://80.lv/articles/building-indie-g ... dern-tech/

First of all, I hope I'll be able to keep this thread on positive attitude, at least until release of H 16 (next week, if I'm correct). After that we can go as usual

I think this is different category of 'sharing'. Let's say there are three types:

1:
Article seems to talk about using HDAs for kind of referencing, perhaps some simple modeling tasks too. Hull of car is one asset, wheels are another so they are finally combined, something like that. Honestly don't know what exactly to say, perhaps Houdini is great here, but, there are other apps for that purpose, too. Exactly where I live and work, real game dev facility is around 20 people, content creator is Max, then sometimes some Blender, Lightwave or Maya. (''real game dev facility'' = which is able to pay for work when work is done). For asset management, usually there is in-house system, or something built on top of existing app or engine. Don't know for any Houdini for that purpose, yet, maybe because no one of these facilities ''fits down'' into Houdini Indie licencing, perhaps the price of full H is too much only for that purpose.
Anyway it could be interesting to see Houdini for character assembly (body parts and props are modeled separately and assembled in H or else), or sort of city/house builder where user is able to combine elements like doors windows or such. Didn't noticed anything about ''character assembly'' in H, yet.

2:
I understood your question as sharing of live operators, all them based on same visual programming system. While this is possible in H, simply I don't see the significant 'scene' available for public, all that after 7-8 years of H apprentice, Orbolt or such. Compare Orbolt and RRray XSI, after all. While relative new Max MCG already shows some ''ICE style" operators for similar tasks, plus there's Fabric for Max, ePhere Lab and ZooKeeper (Ok ZooKeeper is different thing), whatever.

Long story short, 'scene', similar to Softimage ICE sharing, does not exist anywhere else. While there is a chance to see something similar for Max. Personally I have theories about, but let's stop at this point, for now.

3:
budget, advocatus of diaboli POV...
Personally I'm using Maya LT and Houdini Indie because they are affordable, occasionally this fits to common usage of Houdini whenever Maya is (too much) unbearable, where H is a bit boring, but predictable app, good for "indirect " tasks (indirect = using 3d viewport only as display). This is quite good for fun and I don't complain, but, if I'll be able to afford to have Max and available systems for Max, just for fun, I'll be able to forget complete my H story. But this won't happen in near future, so..... would be interesting what authors of these article will do if Santa Claus will come with full Max licences one day, instead of usual Santa Claus stuff, hehe.

[Bullit]

I'll have my sword drawn if they don't have cloth at same level of Cinema 4D or 3dsMax...
-----

My question was more along the lines that if his modeling strategy could be extended for other parts of Houdini, animation, simulation. Let's say i want to add a specific noise - probably something more complex but just for the sake of simplicity- to an animation can an HDA be useful for it?

Btw since i mostly use 3ds Max in my temporary studio job i keep updated on it even if i don't work with last version. MCG is already dead or near it , it never took off due to need compilation, and obviously is less integrated, flexible and complete than ICE. Max version 2017 instead came with a Data Channel modifier that has a lot of possibilities but not surprisingly the GUI is obscure and not friendly. The fact that choose to develop this extensive capability in a Modifier instead of MCG leads me to thing that MCG future is in doubt.

[Mathaeus]

I'll have my sword drawn if they don't have cloth at same level of Cinema 4D or 3dsMax...
-----

My question was more along the lines that if his modeling strategy could be extended for other parts of Houdini, animation, simulation. Let's say i want to add a specific noise - probably something more complex but just for the sake of simplicity- to an animation can an HDA be useful for it?

Regarding Cloth sim, I think there applies the 'Inverse Maya rule'. Because nCloth in Maya is still one of top solutions, perhaps they will not try to compete, seriously. BTW a lot of factory nodes are HDA, user is able to dive in and copy something. Smoke fluids are pretty much modular, assembled of many smaller HDAs, like small noise, big noise, solver, something like that. Now it's about user and task, to make decision, does it worth to go with completely procedural approach. Best way is to try, it seems to be not easy to explain by words. Imho, example of new hair tools in H is looking like simulation used for static hair styling.

Ok back to business, here's another dirty trick that surprisingly turned out well. It joins a lot of overlapped curves at intersections in 2d space, also it performs a cleanup of 'rests'. This is needed here to supply the appropriate input for Polyexpand 2d or PolyWire SOP (here's Polyexpand 2d in action), these two are able to do extrusions along many overlapped curves, nicely joined - if - there is appropriate input, like edges of mesh.
A lot of overlapped curves, that I had with 'small harness system' here, well they are not appropriate input. Curvesect SOP is not enough, it wants two separated curves, while I have around 400. And yeah, 'usual' math to find closest point on line or line intersection does not help there, as it considers 'given' points, while nothing is 'given' here. So, tried a bit of brute force, curve segment based approach.
First step is a sort of extrude to get geometry for ray casting, that's Sweep SOP at start. "Core" of method is ray casting (Intersect VOP in H) from actual to next point, exactly along length between them. In Houdini, ray cast length is defined by 'raydir' vector. Finally point is just moved to position created by ray casting.
The rest of nodes in Point VOP belongs to fixing the possible errors immanent to method, like avoiding to ray cast to another curve, moving the ray cast origin to do not catch 'self' geometry, two ray cast/intersect VOP's to get proper midpoint of that moves, fitting into 2d plane as well.
SOPs after that 'core' 'atribvop_main' are: Fuse SOP to weld points, loop for cleaning the 'rest' points that have only one neighbor, cut to allow nice re-sampling in between intersections, (that's Resample SOP), another welding by Fuse Sop, finally, Polyexpand 2d.

Now I have to admit, that manual welding in some standard 3d app, could be a way faster , but well, this is an experiment.

Another nice surprise outside H, is Blender, and Blender's ability to have empty wire frame, and to import '2d polygons' as plain obj file, and.... to fill all them in single action, if needed. Thanks to our member Wesserbro for point out to that 'empty wire frame' ability, here on forum.

[Mathaeus]

Time for update. Wanted to have a harness system with something more interesting than set of plain stripes, created by sweep and extrude in H. Also, wanted to have elements to fit on existing system of curves, but... without stretching, typical for parametric deforming by curves, and.... all that as much simple. Figured out that each of around hundred curves, have a different length. So finally it's mixed UV and length based, custom deformer by curve VOP SOP ( that's called 'attrib fit mesh'), able to process multiple copies at once.
Not a rocket science to do, while solution still belongs to science fiction in traditional 3d modeling. One 'processing unit' (of around 8, fitting to around 100 curves) is shown in screenshot. It starts with imported 'lego bricks', created in Maya. Usually, start and end are deformed evenly along curve length, while flat portion in middle is stretched to fit along particular curve. That stretched portion is nicely visible on top stripes, I think. While H Duplicate SOP is able to output group for each copy, found more convenient to identify the copies by point numbers directly in VOp SOP ( that's small 'nb points' VOP SOP before final duplicate, just to save the count). Dividing the point number by count of individual copy, so on. Entire thing is still quite easy to adjust, even for now, it is assembled of three 'layers': control points on UV locations, system of curves traversed between control points and fitted to 3d mesh, and final harness system deformed by curves.