early Truespace arrangement model


parametric CFD outlines

Well, it's not really a new project since I have started it over and over in the past like many of my other 3D projects. Mostly stopped it due to time constraints and such. Although, I have become good at creating game sized modelling it's sometime a tedious work when it comes down to curvy models. For one they tend to become big and it's hard to decide the balance and get the topology right. And most great designs tend to have lots of curves whether they are simple, complex or hidden.

Needless to say, Sol Bianca is one of my favorites since it came out way back in the 90's. Sadly, it never became that popular in japan because of it's foreigner cast character designs.

The Sol Bianca is a all curvy design which really hard to create as a low polygon mesh. Looking at it I knew I couldn't create it with foresight in terms of polygon reduction. Foresight is important and save a lot of trouble to fix things. Even one single wrong line can make a mesh look ugly. And it's very easy to oversee one line, escpecially when the mesh gets auto-triangulated during export. This always is the case since I'm using mainly truespace. Now I'm using wings3D to do some work my old TS can't.

Aside from my normal creation workflow I looked at alternatives for this specific project. Nurbs were out of questions due to their uncontrolable growth in meshsize, but I still wanted easy to make with some good control. Imaging Inkscape for 3D. Well, today we got surface modellers that do a great job, but are worse in terms of learning curve and convenience. Also there is a world of difference between CAD surface/solid model and 3D modeller volume model. It's near impossible to get a translation between both easily. Besides, the most easy ones turned out to be too free-handed than allow some accurate designing.

One important aspect was to keep meshing low. Most CAD programs don't have good meshing control or at least not as I would like.
So I ended up misusing a cumbersome but precise and simple CFD program.
Cumbersome in terms of creating it using parametric coding, e.g. mathematical expression. That's right a ship created purely based on math. The good thing is I have greater control over meshing, although, I don't know how it will turn out yet. I have to finish it first before I can see its working...

Nearly finished


but despite all "sure" measurement it seems a bit off:


Measurements and placement are based on line arts and side view screenshots. Two (line art+shot) are 100% in agreement with a corrective factor of 2, and another was off by 7%. So I'm quite confident. Still, I guess I will hall the distance between the bodies, resize thickness to 75% (make the tail shorter).

It's nice!
Not sure why you want it to be so accurate since your source is 2D designed animated starship which never appear the same on 2 frames anyway...
So do you want it to be low-poly model suitable for game, or maybe "full-scale" model with fine details?

AlexT wrote:It's nice!
Not sure why you want it to be so accurate since your source is 2D designed animated starship which never appear the same on 2 frames anyway...
So do you want it to be low-poly model suitable for game, or maybe "full-scale" model with fine details?

Gamers bitch a lot on this kind of things.
Both, that's why I did it like this. I have the option to remesh it to any level of details I want. And I can tweak remeshing for each point or line. It requires a lot of work for small details so I will add them with premade detail parts in a normal modeller. As a bonus I can do CFD/wind tunnel/pressure/temperatur test. The only problem, it seems,the support software is not working correctly under windows.

In any case, it's going to be at least 5k big so it won't do for many regular mods. The Freespace 2 mod can use it, thought.
I will throw in some turrets seen in Sol Bianca2, and JGZinv will be happy.

Oh, well, lessening the thickness caused more inconsistency with the width-2-height ratio. And since other pictures only confirm my measurements I decided to tweak a bit on the tail and nose, while raising it as planned. It look a lot better now. But I suspect the optical accuracy perception lies in the 7% disagreement or error. A bit higher error margin than my usual standard but can't be helped.



I started test meshing of areas I expected trouble due to angles.



As you can see I need to increase control lines and details in order to "guide" meshing in some areas. It's not much of a problem since I foresaw them and reserved some leeway and room for additions. The problems is rather tweaking until things turn out the way they should.
The circular wings are probably more troublesome than the curvy body.

Looks good and accurate so far. Can you shot a few pics with perspective\camera view next time? I always having a hard time getting "feel" of shape and proportions with "flat" renderings...

Done.
animation:




Only need to create surface, tweak the meshing resolution, then export and tweak add-ons..

Edit:


For some reason, the program won't recognize some closed surface and won't create them. Also it didn't mesh some that were recognized.
I have suspected the math library wasn't quite done "well" when some parametric lines didn't turn out the same as in my modeller. Now, I think it has more underlying problems. Perhaps the program has bugs in addition. The bad news is I can't export unless it's "complete".
The good news is I can transfer the parametric information to an alternative CAD program if I have to start anew.
I will just try an older version later and see.

I still hope it will look like this in the end:


Edit:
Problem solved.
I met a new problem. The program aborts loading after only 89 errors. Those errors are minor and don't really matter
It works fine if I do it in one session, though. One session meaning assigning all surfaces/volume at once. duh.

All done and exported.
As it turns out the meshing is more angle oriented than topological line to line. As a result curved surfaces may become edgy between lines.
I can remesh using another resolution but will leave that for later.
Also it appears meshing isn't consistent.
I guess I could use meshlab to smooth things out.
Once the mesh is done, I will start try the reduction process.



See the failed meshed wing?

Looks like it's a pain to do in True space...so much easier in SoftImage -there you just flag hard edge/polygon/triangle where you don't want it to smooth and just press +\- key to increase complexity of the mesh.

Truth is - most of 3D modeling software was never intended for precission modeling unlike CADs. Not that it's very important for animated ship anyway - since there wasn't all that many details you're free to see em as you'd like... I remember when i modeled Roil - what a pain it was, take 3 screenwhots and it's somewhat different in all 3

Well, Truespace is long out of date, but imho the one I can do precision at a way faster pace than any modeller. In most modeller you have to control the views all the time plus going through all kinds of menues just to get precision. With TS and Hexagon you can do it on the go. Imho Hexagon seems to be the ideal TS replacement, but I haven't used it yet. The other good alternative between both world is Rhino with T-Splines. It's not exactly intuitive but has more convenient potential than Blender. Blender & wings3D imho are a pain with their camera control alone. However, wings3d is very convenient to fix the elements. TS is too precise here, hence, requiring sophisticated manipulation. TS can flag faces to increase complexity, but as with most modellers it's not an ideal choice. You still have to fix curvature. The special about his here is adjusted surface sizes depending on topology. This gives smaller and well curved meshes.

The Roil does have many details, you just only looked at 3 pics instead of all pics.

Still I prefer to have the curvatures, placement, and to some extend based on my judgement the details as accurate as is acceptable to me.
As I always say: if you spend some (more) time with a project then go all the way. From experience, if you do bad you are likely going to redo it (over and over). In the end you spend more time in total then when you do it right from the start. I do revise parts of a model regularly but it's rare I start completely from scratch again.

--------------
Anyway, I originally estimated a low level of 5k and 20k for a well meshed model. As it turns out and I'm quite surprised, the high level mesh is at 19224 triangles and the low level at 6000-7000.
Now that I'm optimizing and fixing, I have to say 5k is barely at an acceptable level of smoothness. Doubling the surface details and it would be 20k according to math. So I'm pleased with the size but not really that happy with the meshing. Still the surface division is mostly better than me doing it by (math/3D)imagination and eye judgement. Wings3d is really good at fixing quickly. It's just no good to do the precise curvature. I will have to add some extra smoothness to certain parts. Not sure if I should add all the engine details for a mod. I expect it to have at least around 7k when done. Certainly one of the best meshes at this size.

In any case, for the future I will use this new technique only for quick reference prototyping of a topology. As it turns out I can save a mesh even if the model is not complete.