Biosuit WIP

[Mathaeus]

Time to define the helmet area. Tried new booleans, to re assemble the classic Nurbs modeling style, trims and curves derived from trimmed edges, and geometry created around that. Worked nicely, while it needed a bit long chain of nodes to get a nice, symmetrical curves from 'trims'. Things like Polycut SOP to remove one side, Resample, Ends and Join SOP. Finally had to create an automatic 'curve direction checker and inverter' to get the intersection curves to point always in same world direction, in order to avoid flipping of all stuff created later, like extrusions and so on. It seems I've found some 'dead angle' in method how Boolean SOp creates the 'seam' curve, which inverted by any small re position of boolean operands. Anyway, nothing strange with such operation, and thing become stable after my auto - inverter.
Organic modeling still belongs to Maya, this time this is only a 'body extract'. Rest is Houdini.

Mainly used good old Skin SOP for extrusions, this one creates nicely ordered meshes, with inherited point numbers from cross sections, so they are evenly distributed along "UVs" (btw new Polyextrude SOP seems to creates different point ordering than old one). This allowed easy modulating by VOP SOPs later, while original extrusion is just a plain cylinder. Other than that, Join SOP is nice for combining multiple curves into cross sections, and of course Resample SOP as 'sister of procedural curve modeling' .
Smooth blends, especially on backpack thing, are result of, well, my Mesh Blend HDA, already posted here on forum.
Everything is still adjustable, for now I have less than ten separated geometries. Somehow that 'adjustable' takes more time, every now and then I want to change this or that, but that's about me, not about Houdini.

[Mathaeus]

So that is, probably the end of polygonal/topo modeling in Houdini, this time. For rest of 'high precision stuff' like hair, already have my own routines. Everything else belongs to what I'm calling 'low precision', some procedural for background and such. Here 'low precision' is also a well known for ages, typical procedural crap, like voronoi, reaction diffusion, growing algorithms. Nice looking, but very very hard to sell.

While I finally got what I wanted, somehow it ended in same way as it was in times of H 13:

1:conceptual part, "I'll use this method for that" , getting the main shapes, contour. Everything is great, even it is fun to play.

2: certain SOP, supposed to hold the solution.

3: discovering that mentioned SOP is not nearly controllable enough, by supplied expressions.

4: building a lot before - after VOP stuff around, to get what I wanted.

5: fighting with details, fixing the fixes.

6: figuring out, that from scratch solution would serve a way better, but now is too late to do everything again.

This time it was a PolyExpand 2d SOP for small stripes, had a hard time fighting against crashes on (probably ...) too small distances between source points, trying to avoid its behavior of creating polygons around ends of open edges. Finally had to create several long sequences of groups for "closing" the structure on uv borders, practically one final group for each Fuse SOP. At this point, thing became only "conditionally procedural", prohibitively dependent on result of previous ops - one small error in fitting around uv, caused PolyExtrude SOP to create a wild long extrusions along normals. And so on and so on.

To be fair, half of problems belongs to general 'curse of proceduralism', which is great, fast, non-destructive and all that, when everything is about getting the contours of main picture. But once when it comes to details, price becomes exponentially higher. While H really does not trying to help there, by insisting on more or less classic modelling ops, extrude, bevel and so, for unknown reason supposed to be procedural, suddenly ( while I honestly don't know what could be 'non-classic').

Long story short, when it comes to modeling, app like H could be a helper to Maya with its latest procedural attempt called Paint FX, but not nearly a competitor to concept of Max or SI, as strong but still very 'direct' *and* 'procedural' app, with plugins dedicated to very specific tasks.

Direct modeling is still totally out of industry standards ( that's Max or Maya..), doesn't looking finished. As modeling professional, I just don't care about Side FX escapades with hotkeys, procedures, anything. Where is Maya or Max interaction mode, by the way.

One especially negative moment is feeling that I'm only one of few idiots who is trying to get something meaningful from H, from time to time, while majority is constantly running around step 1...

On positive side, whenever I fall in Houdini 'solutions', somehow I am able to imagine an workable Maya equivalent. No one app is a such strong helper to Maya, like Houdini is.

Anyway, once H modeling part is finished and rest of modeling belongs to predictable, quite linear workflow in Maya (or whatever else), life should become easier.

[Bullit]

Disappointing. I got a lot of feeling that just one more effort from developers could improve things several fold. I sensed it in Softimage, i sense it in some parts of Max...

[Rork]

Well... At least the people at SideFX listen, and implement if it's a true improvement.
Often very quickly.

Keep sending them requests and videos explaining the issue, and it might show up in a next build. ;)

rob

[Mathaeus]

Disappointing. I got a lot of feeling that just one more effort from developers could improve things several fold. I sensed it in Softimage, i sense it in some parts of Max...

Well you know how it was in Mental Ray times, actual version 'was not there yet', but next one will turn the table, hehe.
By the way I'm not complaining there, it's more a sort of research of interesting phenomena of alternative approaches to 3d creation, like Softimage ICE or Houdini. But this time, hopefully I'm not 'directly' tied to Houdini by some tool flowing and used around, which forcing me to 'take side'. Finally I'm feeling free to talk about.
In real life, I reporting what I have to say to Autodesk, and believe me or not, they (that is, Maya team) have ability to ''predict'' what I'll want in area of rigging and animation, few steps ''before'' my mind, for example, complete Time Editor in Maya 2017, or Bake Deformer in Maya 2017 update 3.... well together with a few issues with window flying around but well nobody is perfect .

Never used Houdini for making money, have no plans to do that in future, so, reporting and all that, that's just out of level of my engagement. Except if there is something really horrible, which is not, btw.

[Bullit]

You are complaining when you need to fix the fixers, but in my book complaining is good if justified which is the case.

[Mathaeus]

You are complaining when you need to fix the fixers, but in my book complaining is good if justified which is the case.

Well if you say so

Fixing the fixers, that's quite normal for research and development, but, not really for modeling work for deadline, where someone has to calculate in advance, how long it will takes to finish the thing. In other words, as long as someone exactly knows every step of future creation, procedural approach in app like H is just fine ( procedural approach here = using network editor, only, which is not a rule, just my decision), here this was in case of 'large stripes'. But for small ones, I have to admit, I've started without precise idea, and ''goddess of procedural approach'' punished me.
If I did everything in traditional way, it will be to hell boring, but in predictable time, that's it.

Nothing new by the way, that's the same genus proximum as it was with ICE. However, imho there is significant differentia specifica: even in early ICE times, people started with building the routines, systems, real solutions for real problems, plugins. And of course it was SI as state-of-the-art strong app, able to hold and fix whenever is needed, which Houdini as traditional app is not, alone (with all due respect). And.... for my taste, always inappropriate reaction of Houdini community. Instead of building the real solutions for real questions, there's 'forwarding' to tutorials (usually something general and not adequate), or even worse, pushing the poor people into secrets of vex coding. They never heard for division of labor, or what.
In short, unbelievable ability of Houdini community to always go back into initial state of cultural revolution, that's most fascinating part.
Side FX, well just like any other developer, they are doing what they can do - while old app has a lot of 'simply can not do'. Things like kinematics, real time evaluation, most likely will *never* reach anywhere close to Maya, for example. Despite the 'listening' BS.

[Bullit]

Yes the deadline is the problem.

In short, unbelievable ability of Houdini community to always go back into initial state of cultural revolution, that's most fascinating part.

Yep reminds me of some in Blender community completely oblivious to absurd Blender interface.

[Mathaeus]

Yes the deadline is the problem.

In short, unbelievable ability of Houdini community to always go back into initial state of cultural revolution, that's most fascinating part.

Yep reminds me of some in Blender community completely oblivious to absurd Blender interface.

Yeah, something like. Perhaps Side FX licencing style, with that strange commercial-which-is-not-really-commercial, helped to gather a lot of people from ''undefined borders'' of 3d world, outsiders and beginners of any kind.
While Blender community is really huge, today, there's diversity, different branches, thousands pieces of great work.
Sometimes, we likes to forget, that Softimage ICE, as a system who was able to be accepted in so, so, so much wider range by artists, than Houdini will be ever, that is, there are entire categories of artists able to take H only as an brutal offence (including me, when I'm animating).... well this ICE experience gave to us a lot of info and knowledge about audience, which seems to be unknown for anyone related to Side FX, from developers, people on forums, anyone. On another, bad side, never enough of 90s style Linuxoid arrogance, so we have that self-disgracing H thing rolling around, for years.
In any case, one day when we will see something like Mootzoid stuff for Houdini, someone who is able to live providing the real solutions for real problems for individuals *and* studios as well, that will be a sign of ability to survive out of a huge facility zoo. Nothing of that until now....

[Mathaeus]

Back to real bussines....



So... raygun, as an essential part of such imaging. I think it's interesting how this one evolved through years. First design belongs to MoI (MoI is only NURBs modeler, no polygons there) as a part of re-building the
Nurbs skills. While MoI imho offers one of the best ergonomy on planet, drawback for such work is modeling style similar to Rhino concepts from 90s. Everything is around somehow deformed Nurbs single patch (loft, revolve, so on), and trimming and blending over that. Has no built in system of smooth stitching, like t-spline. That makes it great for adding the small details, cuts, but unacceptable hard for building the blob style, long transitions between volumes. Also, every hard cut has to be beveled manually. Finally, DCC apps, all except Maya, are not good when importing these trims over trims, blends and such. To be fair , MoI has great mesher, but then I'll have to deal with huge meshes, custom normals and whatever.

Next one is kind of retopology in SI, now only by polygons, suitable for subds.

And last one for now, is re - design in Maya. Nothing that can't be done in SI, but my SI is 8 years old XSI 7.01, perhaps there's time for change. Found these 'wings' unpractical, wanted entire thing a bit more smoother. Basically modeled it in same way as in SI. There's always what I calling an 'etalon', base main surface having no details. Rest is snapping to etalon, Shrink Wrap in SI or Maya Live or Maya Shrink Wrap. After some adaptation, found 'Maya way' a bit more handy. Good old Maya Live is a kind of snap, not an operator, it's a bit easier to turn it off / on. Tip: setting the Transform Constraint in Tool Settings to 'Closest Point', allows Maya live to nicely snap to target boundaries. Another Tip: double click on tool icon (move tool or else), opens the Maya Tool Settings window.





And few more tips.. If you don't like Maya radial Marking Menus that much, good news is, it's possible to avoid them to some level. So, for getting the Maya Marking menu to behave in a way ike in pic, procedure is:

- find appropriate MELs in scripts/others and scripts/startup. List in pic, are MELs responsible *only* for poly modeling, component selections and transform constraints (what you getting by holding WER keys, once certain transform tool is active)
- *copy* them to user/prefs/scripts folder. This makes them 'preference' scripts, Maya will load them instead of factory ones. Also, this makes it easy to go back to defaults if needed, just remove them for user/prefs/scripts folder.
- comment out, or delete something like -radialPosition "N", or -rp "W" ("N" or "W" means north or west, so could be something else). This tells to Maya, to just collapse the radial menu into list.
- do it in Maya Script Editor, at very own risk of explosion, Maya wildly running around room, anything. In case of new version or update, it's possible you'll need to do it again.



Selections, components....

In SI, we have a magic TUI keys for component selections. For decade, I've used only these three plus another custom three for converting the component selection. In Maya, that's more 'granular' process. So, to get something similar, I'm using custom sequences. It's really easy and handy to create them in Hotkey editor, just go to Runtime Command Editor and create them. Maya admirably loads them once they are saved. This is equivalent of T key in SI:

Code: Select all

selectPref -useDepth false;
selectPref -paintSelect 0;
strsTweakMode(0);
setToolTo selectSuperContext;
selectMode -component;
setComponentPickMask "Point" true;
setComponentPickMask "Line" false;
setComponentPickMask "Facet" false;
setComponentPickMask "ParmPoint" false;
selectMode -component;

It's a bit long list, but I think it contains the all possible combinations for 'safe' switching, in case you have appropriate scripts for polygons or edges. Consequently, permanent tweak component mode is something like this:

Code: Select all

setToolTo moveSuperContext;
selectMode -component;
strsTweakMode true;
strsTweakModeDoIt(1);
selectPref -useDepth false;
setComponentPickMask "Point" true;
setComponentPickMask "Line" false;
setComponentPickMask "Facet" false;
setComponentPickMask "ParmPoint" false;
manipMoveContext -e -ah 3 Move;

Not sure is at all needed. By default, tweak Mode in Maya is holding the tilde key. Great part of this, is easy switch from manipulator to tweak, there's great possibility to tweak the SRT, too. Not so great part become with snapping, one has to hold tilde *and* snapping key (that's great if you're octopus, anyway). One special moment is 'inheritance' of move tool if moving along one axis, tweak mode will stay in that axis. " manipMoveContext -e -ah 3 Move; " at end is just to prevent this 'inheritance', setting the move context back to all three axes.

Last tip: there are ( at least ) two types of "component modes" in Maya. One is 'true', what you'll get by F8 or scripts like these. Another is kind of mixed mode, object + component (title bar shows nicely that is primarily object mode), this one is called by main object marking menu, and many other commands. From my understanding, purpose is to allow setup of things like partial shrink wrap (some points wrapped to certain object). But, it's also confusing for users of "cleaner" interaction like SI. To get out of that, object + component mode , user has to press F8 toggle twice, or, with scripts like these: first to go into component mode, then into object mode.

[Bullit]

I modeled with nurbs in Rhino and Alias StudioTools - not sure if it was called that way at time of that name came later but it is how remind it. It had a radial menu that was essential to work with . I also modeled Nurbs in Maya but was just one work. It was Maya 2.0 about year 2000 i think

So how to model with nurbs have no secrets for me. I also modeled furniture in Solidworks so also know to model with solids kernel kind- well it was more store reception desks and like so not serial production.

Regarding Softimage your 7 version don't have the M key isn't it? i think that is great tool. There were also some ways of maintaining curvature that were good in some later update but i never used that.

[Mathaeus]

Regarding Softimage your 7 version don't have the M key isn't it? i think that is great tool. There were also some ways of maintaining curvature that were good in some later update but i never used that.

If you mean thing that's showing a three little on screen buttons, yeah XSI 7.01 has that, feature was introduced in XSI 5, I think. Anyway in XSI 7.01, loop - ring select is old fashion alt+ something, not double click.
Btw if you'll want to do nurbs modeling in Maya again, perhaps everything is same as it was 2000 .

[Bullit]

Yes. Nice you have the M button there were other updates with selection later with double clicks.

already at that time was outdated compared to Alias and Rhino.

[Mathaeus]

Yes. Nice you have the M button there were other updates with selection later with double clicks.

already at that time was outdated compared to Alias and Rhino.

Honestly never used M thing that much, or, used it in reduced way similar to older tweak tool. Today I think every app has some ''edge slider'', Maya has a bunch of related options, but, also not that much of benefits when it comes to precise work. Simply don't know for anything better than old fashion snapping to reference surface. Only that's substantially new, imho is that the same process is called re-topology, today.

[Bullit]

For me is important because i want the creative process of designing in 3D instead of 2D.

[Mathaeus]

Small update, now it's a bit boring phase of adding more and more details, slowly getting into final shape. As usually, have a mix of Maya and Houdini work, some things passed a few forth - back stages, like current ''backpack prototype'' - draft is Houdini and my Mesh Blend HDA, later it was -topology in Maya, perhaps it will pass several more such phases when adding details to it - somehow found Houdini great for prototyping such things, while Maya is holding the all final work.

Just to share a network used for deformations. Everything related to this character is created in static pose, later is deformed by pairs of meshes, exported from Maya. So, static pose and mesh deformed by Maya rig at certain frame. Because it is Maya LT without Alembic, and final is still picture, I'm using good old, simple .obj files.
Of course I've tried factory SOPs before, like Cloth Capture or Point Deform, but they weren't suitable. Cloth Capture was accurate by prohibitively slow, even on short sampling distance, Point Deform was too ''loosely''. Both were not comfortable for blending.
Network is showing a sort of mix of polygon and point reference frame (in ICE naming) - XYZ Distance VOP is a kind of ''half" of ICE "get closest location'', telling from which primitive (polygon here), to sample the meshes, by bunch of ''primitive attribute" VOPs . (''UVs'' in ''primitive attribute" VOPs are determined by polygon edges, it's not what is usually considered as ''UV"). Important part are Normal SOPs before - once they are creating the normals on points (should be just that, on points), N in ''primitive attribute'' VOPs is interpolated, allowing a nice, smooth deformation. In case of not existing N attribute on points, or, in case of using ''primitive normal'' VOP, normals are *not* interpolated (that's close to ''Surface GeometricNormal" in ICE).
Such network allows me to mix the smooth deformation for clothing together with rigid deformation for eyes, helmet, backpack, ray gun. All that using only one skinned mesh in Maya. In case of rigid deformation, it's just one constant vector value used for sampling by XYZ distance VOP - on some place where just one Maya deformer has a full weight, like chest or head bone.