Model Editor

[uzurpator]

I've been playing recently with Lego CAD programs and it struck me, that it is a very nice way to actually build models for TE.

Basicly - we would create an editor that would allow to build models out of 'bricks' - small cubes - that would be later converted into regular wireframe. Building this way is extremely fast, and does not require any texturing skills - any TTD artist could jump right into it.

The model below took me 5 hours to build - despite me having _no_ skill in modelling. And this model is very detailed. For our game 1/3 of this would suffice.

Waddya think? And as a bonus - any takers for thet ask of building an editor?

[DominionSpy]

That is an amazing idea! I like it.

I also have hardly any skill with 3D modelling, but I've used this lego CAD program with ease. Lego is very versatile and making changes to a design is really quick. Plus there are sites with well-known trains in lego form which we could copy (with the author's permission of course).

[Hyronymus]

But there are limitations. You can't rebuild every design in LEGO, there simply aren't sufficient brick types. So how does one solve shapes that can't be re-created in LEGO bricks?

BTW: the model looks fine indeed.

[Mek]

I've been playing recently with Lego CAD programs and it struck me, that it is a very nice way to actually build models for TE.

Why would the coding team actually worry about creating models? isn't something completely unrelated, and if it should be discussed it most certainly should not be discussed in the Coding forum i guess..

[Steve]

Having a set of little bits that people can add to their trains would work, but I don't like a full brick building setup. For instance, you can reuse bogeys.

[uzurpator]

DominionSpy:

I was'n thinking about using actual LEGO models. But more in the lines of using one of the lego CAD programs as a base to build our own editor using the same principle.

For starters - we are going to use different scale then legos.

Hyro:

The question must be asked then: What is the proposed _view_ of the game. Or more specifically - do we keep the TTD way, but port to 3d (think Warcraft 3) or allow fully freeview (think Railroad Tycoon 3).

If it is the WC3 (as I kind of assumed) way - then the models will be _tiny_. With 4 pixels/meter we are struck with trains similar in size to those in Locomotion. With such small size building full-blown models is a waste of time. While building them of small atomic 'bricks' (or predefined collections of bricks) will suffice to achieve sensible lod. As a bonus we will control polycount, create skins automatically, and will be able to control scale of the models we get.

Also - using atomic elements will assure that models are simple to build. Frankly - it won't differ much from drawing the 'side' view in TTD. With a bonus of not needind to do shading/texturing because these will be handled by the engine itsself.

mek: While making models is not of our concern - model format, and editor is. If modifying a LEGO cad editor was easy enough, then we could have our own collection of models in a matter of days. (Granted - someone would have to make a converter to a wireframe, but that is quite easy task)

steve: what is the problem with creating a vehicle out of several models (built of atomic elements that is), pantos, bogies and - say - steam loco running gear are candidates to prefabbing.

[Steve]

Whilst they will be quite small in the general view, will there not be situations where we can view them close up? Like if we want to have a good look at a station. Or the pictures in the train builder window.
Would you actually build 2 models for this purpose, 1 high def, 1 low def, or would the engine be able to simplify the high def models?

[PJayTycy]

Would you actually build 2 models for this purpose, 1 high def, 1 low def, or would the engine be able to simplify the high def models?

I think we should aim for 5 models. Where the lowest detail one is only used when the object takes +/- 3 pixels on the screen and the highest detail one is only used in very close up.


BTW, uzurpator, I'd favour completely free camera movement, but if you think it's too much work / too difficult for a small volunteer group like ours, then drop it for the first few releases.

[uzurpator]

I think we should move this to the Development forum.

Anyhoo - Answer me - how big zoom is big zoom.

4px/meter will translate to 104 pixels
on 1024 px screen that is 10% of the screen width. That is quite close. It display 16 tiles in that resolution.

Take a peek below - that is how 4px/meter models will look at native resolution (more or less). You can make closer zoom without big detail loss.

PJT: You make one model, and let vertex shader units on the graphics card to do the lod.

Zoom or free camera is not a problem. But if we want to have _really_ close zoom (what for?) then to get decent model quality prepare to _wait_ until we get enough of them.

Also - with an open source project we want to make the model-making as easy as possible. Atomic modelling is easy, but suited only for small models.

[Steve]

Looking at that screen, that is probaly the closest I need to go. You can see the train nicely within the town, and then you zoom out a bit to make the tracks and stuff.

I think it's important we don't sacrifice on quality now, so that we don't have to remake models in the future.

[Purno]

This is a great way indeed, this way I would even be able to help making models!

But there are limitations. You can't rebuild every design in LEGO, there simply aren't sufficient brick types. So how does one solve shapes that can't be re-created in LEGO bricks?

Well, someone with 3d modelling skills could make additional bricks we need.

[Purno]

I think it's important we don't sacrifice on quality now, so that we don't have to remake models in the future.

Even with bricks we can have different qualities.

Look at these two pictures for example:

Small quality
Large quality

Both with the same size of bricks (LEGO)

[Hyronymus]

I don't call both quality to be honest. It's merely an 'as close as possible representation in LEGO'. I don't understand how you came with the entire LEGO idea anyhow, uzurpator . If we were making LEGO Trainz I can understand it easily but we're not doing that.

First of all we're trying to make a transport simulator that has road vehicles, trains, ships and aircraft. My second point is that if all other transport sims can have decent 3D-support why should we stick to an incomplete 'solution' such as the LEGO editor. My third and last point is that we said almost from the start that we would work with real 3D models that can be loaded/edited/produced by the 'standard' 3D software kits.

[Purno]

I don't understand how you came with the entire LEGO idea anyhow, uzurpator

Because it's a very very simple way to make models. Even n00bs can make models this way. (like me)

First of all we're trying to make a transport simulator that has road vehicles, trains, ships and aircraft. My second point is that if all other transport sims can have decent 3D-support why should we stick to an incomplete 'solution' such as the LEGO editor.

You can make it as detailed as you want with bricks. (don't have to be LEGO-bricks)

My third and last point is that we said almost from the start that we would work with real 3D models that can be loaded/edited/produced by the 'standard' 3D software kits.

You could develop some tool which can convert those bricks models to normal 3d models, since every brick is a 3d model itself.

[PJayTycy]

PJT: You make one model, and let vertex shader units on the graphics card to do the lod.

Ah, I don't know anything about vertex / pixel shaders...

The most important reason I proposed different detail models was to reduce the memory load on the graphics card (ie: smaller models take less video memory).

[uzurpator]

Hyro.

1. What graphics cards eat is triangles and textures. Any model, no matter how built, will have to bo converted onto triangles and textures.

The problem is that working with raw triangles and textures is painfully slow, that is why you want to avoid it if you can.

But since the models will become triangles sooner or later - nothing will stop you from editing them in Maya or 3DSmax.

2. You can create any model with this technique.

3. The third thing (and this also is to Steve) - accustomise yourself with term of: 'error of discretisation'.

Basicly - the smaller the object, the less triangles you need to achieve full possible detail.

Take a look at the picture below. The full scale br 103 is about 180000 pixels. The upper one is 15% of the original and uses 4050 pixels.

By using 15% of size requrements for mesh dropped to 2.25% of original.

The 15% image is closer to the zoomlevel of TE (if we keep the TTD-like view) - which means that using the bricks we will create models of _maximum_ possible detail. But even with double that view there will be no significant loss of quality.

And please - I'm not talking about building models out of legos! I'm talking about using the principle behind the Lego Cad programs to build our models. Basicly I need precisely 7 elements to build exquisite quality models of every object in the world.

PJayTycy:

Speaking of view. How about limiting the view cone.

Say - camera angle from 90 to 35 degrees (stright down to heavily slanted), zoom up to certain level and free rotate.

BTW - more models per LoD actually increase memory needs since you need to load 5 models instead of one.

[Hyronymus]

What's the point in designing stuff in program A and refining it to meet a desirable result in program B when program B can also do what program A does?

[uzurpator]

If we set T as amount of time, then if program A yields Y results/time and program B yields X results/time and Y(T) > X(T) then using program A is justified.

[Hyronymus]

I disagree, time is not our concern namely (as someone else mentione before). It's 100% up to the graphics artists to decide how much time they're giving modelling, it's just not up to us.

[uzurpator]

I disagree, time is not our concern namely (as someone else mentione before). It's 100% up to the graphics artists to decide how much time they're giving modelling, it's just not up to us.

Really? Good - from now on I develop in machine code...

Besides - We already have two artists (me and Purno) who like the idea...