Yet another AI model standard probing as of 2020 on the military point of view

[integral]

Hi,

Here are some questions I have addressed to Military AI enthusiasts about modeling standards as of 2020, but including all previous Flight Simulators up to FS9 (which I still use because low end computer) I'm currently back in FS AI modeling and primarly focused on eye-candies.

The questions :

1) Antennas and Domes : a) generic model or get them all ? b) use alpha or hide using title + XML (TenkuuDS way) ?

2) Cockpit windshield : a) Translucent or opaque textured ? b) If translucent, basic cockpit components has to be modeled (seats, panels), but what about pilots ?

3) Ground handlings : a) Tow/push back ? b) Tires chocks ? c) Remove before flight ? d) Opening doors ? e) Stairs ? f) Engine/flush/exhaust covers ? g) boundary cones ? e) Ground power units ? f) Refuelling truck ? g) Others ?

4) Engines : a) Model the inside (fans) ? b) Adaptative intake/exhaust for fighters ?

5) Gear bays and tires : a) Model the inside ? b) Rouded tires or flat sided ? (Some aircraft have very distinctive rounded tires, like the IL76)

6) Overall modeling accuracy : I'm on the Erez-like side, use real side/front/top views pictures with as less distorsions as possible to begin with. 3 Views, even from the manufacturer, contain tons of errors, wrong nose/tail/cockpit outlines, misplaced antennas/engines/wingtips. My assumption is, someone, at some point (me for instance, or Erez back in the days) will get the resolve to build his own rendition of the already existing AI model, to fix this and that, and raise the standards up.

7) Texture layout (considering the fact I'm still targetting FS9 with the 1024px limit - sorry) :
a) Larger fuselage/tail and engines texture ? Like usual commercial airliner, splitting the fuselage in half whenever required.
b) Top fuselage mapped appart ?
c) Bottom fuselage mapped appart ?
e) Loadings : let's take an F16, with one hundred loading variations : should I try to squash all missiles/bombs/fueltanks in the main texture or use another one ? (more drawcalls)

8 ) Flexing wings for large aircraft, KC46 for instance ? What about a 737 (a P8 ? probably not ! rather 200s and 300s with domes)


Side notes :

I want to make one thing clear : I do love and use existing models, I do enjoy simflying with them (AI spotting), and I'm extremely grateful to their creators who had put lots of efforts and dedication to bring those addons to the community. I'm in no position to criticize their works the more since I haven't given back anything than an outdated scenery, not even a single AI model so far. Yes, it appears asking about standard implies "I'm not satisfied with what exists", and I can't deny that, I do have wishes, to be honest. But that's just me in my yet another attempt to resume modeling and hopefully release models (at least one to begin with).
Why am I asking is not to start bashing on any model or anyone, but to get other points of view about what trying to push accuracy that far implies :
a) modeling undoubtly takes much more time
b) multiple versions of the same aircraft by different modelers, meaning redundant works in AI ressources, repainting and addon installing.. A mess and wasted efforts : in a way, yes, but on my end, no.
c) some modelers may find the standard too high to get them started in projects (while I highly doubt that)
d) yet again another AI vs UI features ? Is that really an issue ? (I remember back then high polygon counts or flexing wings which are pretty common today). I don't think AI models should be that "basic", nor asking other modelers to comply with any standard, but I do enjoy myself when I'm impressed by the precision and accuracy of some models, flirting with UI standards, and I can model that way and please myself even more.
The above are the concerns I have right now. I'm half, "just model what you want and release it as-is at Avsim and don't mind what happens", and half "maybe, I should share and discuss my thoughts to add the features the community is expecting" as for instance I'm not really interested in removing eye-candy (rather adding as many as I'm willing to), but I have no plan so far to embrace the translucent windshield - that is an example of something I may reconsider, hence, this post. Should I remove something to have that instead ?


Last : what's the plan all of a sudden ?
Time ? I do have now, since I have full control over my part time (changed employer)
What can I do ? GMax, Paintkit (PSD), XML
What can't I do ? FDE and CFG tweaking.
My goal : Enjoy my sim, make models I like, I would love to share them IF I find out a way to get help about FDE/Config set.
Free or Payware ? No plan for Payware.
Which FS version ? FS9 only, again, I'm very sorry. I can't FSX native because I don't know how to, and, potato computer : can't get started in FSX/P3D (I do have FSX but P3D/FS2020 are a big NO, quite expensive and would not work anyway) However, I have no restriction whatsoever for anyone willing to port my hopefully future AI model to FSX/P3D, anyone can do it and release to the public as long as it remains free, this remains valid even if at some point any attempt to get in touch with me fails (if I die for instance. The point is : should you get my written permission to convert any of my AI models and release to the public : NO, just do it, I'll undoubtly use the result of your work once I can FSX or P3D anyway...)
Any plans on beta ? Since I announced my intentions, probably at some point. Where/Who ? My assumption is MAIW is interested in military model variations, while AIG seems appropriate for airliners. FSDevelopper may also be a more generic platform. Anyway, aren't we mostly on all those platforms ?
Which model(s) ? Not a single one in particular. This is a 15 year of modeling reboot (almost like FSPainter-Missi), the choice is wide. Any models set completed first comes first, hopefully. However, feel free to suggest models and if it/they happen to be ongoing dev among my Max files, well, why not ?
What about planning and release dates ? I have NONE and no intention to predict some. Pressure kills motivation, I have no plan in complying with deadlines.

Thanks for reading.
Have a good day.

[Firebird]

There is a lot in there and although I have read it the chances are that I missed some stuff so I apologize in advance.

In simple terms, MAIW was formed so that people that wanted to produce military AI but couldn't do everything themselves could assistance in completing projects and/or assist others in completing theirs.
If you look at our packages I doubt if you will find one that is a solo project. All have collaborations even if one person is named as the primary.
So from your perspective whatever you need help with you will find someone that can create something for you. As for what you want to produce, we learned years ago that modellers are far more productive when they have the freedom to choose their subjects rather than constrained by others.

Aircraft never modelled before, or new models of subjects that were created years ago, we don't mind. Current aircraft or retro, we don't mind. Me personally - I fall into the school of 'if they build it it, I will install it'. I think most people are.

We do not restrict people on what tools they wish to use, again anything that makes a modeller happy is best.

As to the standard of models. My guess is that you are no different to any other user in that you would prefer to exact replicas of each model but we also realise that for somebody to create every single sub-version of, for example, the C-130 would involve so much work that it would be counter productive so common sense by the modeller is perfectly OK. Each subject would be different according to how many variants there are. It is perfectly acceptable to ask advice of either individuals or the general population as to what they would like to see. The big difference with mil aircraft is of course weapon fits. To be honest whilst everybody would like to see every possible fit done it is not practical. A good example of what I mean here is the JYAI Tornado released fairly recently. The more basic models that were required meant that some fits had to be left out or we could possibly have ended up with more models than Tornados actually produced.
When you announce what you are doing you will find that quite often people will ask for this or that. We always leave the final decision as to what is produced to the modeller.

One other key thing is that we do not restrict modellers to be exclusive. Quite often they do become that way but we do not insist on it.

We do not insist on announcements like 'I am building xxx' nor do we insist on development shots or statements. We know that a lot of projects that start do not get finished for one reason or another. We leave up to the modeller how and when he wants to make announcements. We understand that part of the enjoyment is the 'wow' factor when screenshots are posted.

If you want to know more about what it is like working in our framework then maybe I would suggest contacting John Young via pm.
Not only is he also a GMax user but I know he was initially not sure about doing anything with MAIW as he was part of the ACG. He would be able to tell you the good things and not so good things about the org that he found by experience.

Hopefully, you will want to be affiliated with us and we look forward to working with you and benefiting from your projects.

[John Young]

Too many questions in one go to be able to respond to in any detail I'm afraid.

Standards obviously vary, depending the point of the learning curve the designer is on. You could explore my standard and practices (all Gmax), or anyone else's for that matter, by downloading, installing and looking at my recent models - C-130, Tornado etc for FS9 or FSX/P3Dv4. Alternatively, just browse through screenshots on the forum.

We've moved on quite a way from the start of MAIW when polygon counts per aircraft were around 5000 mesh triangles. With more powerful computers and graphics cards these days, 5000-10,000 polygons is quite workable with careful use of LODs. My practice is to use the polygon budget fully, making the aircraft as realistic as I can within that - so yes, transparent cockpits with interiors, instrument panels and pilots with animated heads, bevelled tyres (sorry tires) and cut out wheel wells, at least in the early LODs.

I would be cautious about comparing models, not just because designers will be at different points on the learning curve and in different times over the past 15 years or so. Obtaining reliable, sufficiently detailed, 3-view drawings is not usually straightforward. You have to live with what you can find and what you can find is not always consistent. Avoid importing photos to trace. Accuracy depends on the camera position (they need to be square on and taken at the same datum level as your Gmax screen) with lens distortion minimised. Published data is not always straight forward either. Working out what length means (does that include the pitot tube on the nose or the sweep of the horizontal stabiliser past the tail) is usually a challenge. Height too (is that with a compressed undercarriage and at what pitch) is often more of a puzzle. In the end you have to put all these things together and find the most consistent option to run with. Other designers of the same model will interpret differently. Who's to say who's right?

I've always found that designing for MSFS (scenery or aircraft) can't be done with solo self teaching alone. I worked with fellow scenery designers at ACG for several years and we learned from each other. I eventually found myself needing to design some static aircraft for a particular project. That led to the rapid discovery that I could fly them as AI quite easily and an immense sense of refreshment and a change of direction that I have not looked back on. It was a natural choice to gravitate towards MAIW because the resources and expertise for military AI are here. It's been great to be able to continue to learn, progress and to use the community as a sounding board to try and develop what is likely to be used.

If you understand Gmax then you have a great advantage. I would just get stuck in with something not too complex and wait till you hit a barrier. I and others here can help when that happens - you only have to ask.

John

[integral]

Thank both of you for sharing your thoughts. There are many relevant points already to begin with. I'm sorry, my questions are indeed too broad and should be addressed on each specific aircraft model rather than on a global scale.

I can briefly pick some points :
Yes, I do have interests in aircraft never modeled before, will keep that in mind, but on existing models variations we happen to only have placeholders aswell (FAIB 747-200GE for VC25A, AIG DC9-30 for COMAC ARJ21-700, etc.)
Announcement didn't work that well for me in the past, it ended adding unexpected pressure. On the other hand, recently, gathering materials (pictures mostly) about a dozen units of a particular aircraft per day helps me way better, hopefully that would do for the time being.
About being affiliated, I understand, I'll happily discuss about the fun once I have someting interesting to share
On GMax, I model in Poly mode and create the triangles point per point. The most masochist way, I know, but, strangly, that's how I'm most comfortable with. My weakness is nose cone modeling (not the window frame, precisely the nose shape, tried the A320 and failed, the DC9 aswell... side and top views are good, but once in 3D mode, it's not like the real thing )
Comparing models, I understand the concerns, I'm not comparing existing models I build my own 3 views by piecing together dozens of pictures (quite time consuming). As for precision, I think 1% error or less in length/span/height is not a big deal, what matters for me is the shape and proportions, but yeah, I do make errors, most of the time anyway, nobody's perfect.
Pilots ! Can't get them below 500 polys... per pilot. I would give it a try once more, but I'm not confident I'm good enough to get something ressembling a human rather than a potato.
I've spent countless hours in trial and errors, like any of us. But I can do things today thanks to the community, starting from Project AI forums, AI Aardvark, FSDevelopper, MAIW, AIG and AVSim forums. The mirror trick in GMax, the CTD caused by duplicated material names, tools like MDLCommander, FSRegen, explanations about drawcalls, how to create LODs (the MS tutorials are worth reading but quite scarce on explanations)... just to name a few. I do agree, I can't go that far alone.
I think I have one XML trick that may help us getting dozens of payload variations with only a couple of models. If it works, I'll get back ASAP to detail the thing here. And yes, that's right, we learn everyday

For the time being, I'll keep going on modeling what I can, it's still a long way to go, hopefully get one model to a beta as soon as possible (while managing the motivation), eventually asking for help when I hit a hard wall.

Again, Thanks.

[John Young]

I model in Poly mode and create the triangles point per point. The most masochist way, I know. I build my own 3 views by piecing together dozens of pictures (quite time consuming).

That's probably why you can't get the mesh right then for all the reasons I've said. If you are working in polygon mode, you are not working in triangles, but rectangles, I always model with polygons because I find mesh mode (triangles) too confusing when there is a lot of detail to pull around. Start with a 20 sided segmented cylinder and select each line of vertices and stretch or squash them to the shape in all 3 views, working from a consistent drawn plan not one cobbled together from photos.

Also, learn how to slice selected planes without disturbing the adjacent mesh. You will find that invaluable now to further sub-divide the mesh and especially later when you tackle windows and gear wells. You also need to know how to keep the mesh smooth by collapsing or welding overlapping vertices and by using smoothing groups.

The Gmax pilots in my Hawkeye model are each 515 mesh triangles in the first LOD and 70 in LOD 4. No pilots in LODs after that. It's what LODs are for, as I said.

You are trying to think through too much in one go. Just get the basics right with the fuselage in the first instance and move on from there. Eat elephants in small bites and worry about everything else in stages as you progress.

If you would like me to check a stage for you as you progress, I'll gladly do that. That might save you trundling on into later stages with practices that can be improved early on.

John

[integral]

Thank you for the tips. I'll give them a try