Correct method for importing models?

I have been finding some free models in .obj, loading them in DeleD and exporting them using the Multiverse Exporter(NOT the COLLADA). I then am able to go directly into the world editor, find the model and place it in my world.

Is this the wrong method? I read somewhere recently on the wiki about converting to .dae and making sure the textures are .dds. I haven't been doing that, but it lets me import the models.

Is this simply the most efficient, economical way for the engine to better handle the models? When I get to actual production phase, I would like to be doing everything the best way possible, especially preparing ahead of time, for possible future upgrades. If I make sure LOD is programmed just the way I prefer, it will make it easier to swap out to better graphics later, without crushing the engine/game under the weight of inefficient or un-economical memory usage.

In my experience, the Collada exporters from a variety of programmes, vary so much, this is reflected in the results. I also use the same workflow that you do, with the twist that I can apply/use DDS formatted textures in DeleD, then export as Multiverse format, albeit DeleD doesn't render them. This is only a pain if I create the model in DeleD, I use say a JPG format version, then once UV mapped, change them for the DDS version.

If you have 3DMax or Maya, there is a OpenCollada Exporter Plugin that works very well with most programmes (not the native one, as this is borked).

I suspect that DeleD may in the near future, support DDS format.

I think I understand. DeleD doesn't natively support changing the textures to .dds, so you found a work-around?

I will have to look into this. I haven't advanced far enough with my own model-making to use UV mapping at all, but I found the free, already finished models, with their own textures and I import them(from .obj) into DeleD and the textures show up when I do that. Then I just immediately export using MV exporter. Is that keeping the textures as .jpg or something other than .dds, then?

I found the wiki entry and it seems to highly suggest changing textures to .dds.

Thanks for the help.

well actually, I stick with JPG formats, unless I'm using a alpha texture, then I use either TGA or PNG, when exporting from DeleD. I do this because I can UV map everything in DeleD and know, when I change to a DDS in the native programme, of the same resolution, it is going to workout okay.

Oneday, things won't be so complicated, as there seems to be yet another format that says it will cure all ills, just around the bend.

Thanks for the further information. I still don't fully understand(The light just hasn't turned-on for me yet), but I realize a lot of my confusion comes from me, simply, being new to all of this. I have been learning fast, though, and I will wait until I do further testing, before moving forward with too many more questions.

I'm slowly inching closer to having a mock-up done that will let me look at, test, compare and figure out further ideas, problems and directions I either should go or want to go.

For instance: I will be trying to figure out if it looks better and runs better to have my own separate LOD models programmed in for LOD, or if it's quicker and better looking to just have the automatic LOD function.

jstrat74 wrote:

Thanks for the further information. I still don't fully understand(The light just hasn't turned-on for me yet), but I realize a lot of my confusion comes from me, simply, being new to all of this. I have been learning fast, though, and I will wait until I do further testing, before moving forward with too many more questions.

I'm slowly inching closer to having a mock-up done that will let me look at, test, compare and figure out further ideas, problems and directions I either should go or want to go.

For instance: I will be trying to figure out if it looks better and runs better to have my own separate LOD models programmed in for LOD, or if it's quicker and better looking to just have the automatic LOD function.

That is something that is part of the rendering engine. You can however setup Billboards for distant elements such as foliage etc, if the rendering engine does not support paged geometry for instance.

Some rendering engines support Mipmaps and Atlas Textures.

With mipmaps, these are normally created either at import, or pre-rendering stage at the client end. These are a series of diminishing resolution copies of the original texture, used to apply on landscapes, foliage, in fact any element as it becomes further away from the viewer.
Wiki on Mipmaps

Billboards are single polygons that have a texture applied, that always face the viewer, no matter where they are in the scene. It gives the illusion that a trees for instance is always cylindrical/3d.
Billboard Ogre Wiki

Atlas Textures are a single texture that contains several different textures used on either a single model or many models. If the several models are held in memory and displayed for long periods, then having a single texture holding several other textures, saves memory. However it can be a waste of resources/memory, if only parts of the texture is used. Another disadvantage is the limited resolution holding several textures on a single texture/canvas.
Here are some examples
And a Wiki explaination

Thank you for the information. It is a lot for me to absorb, but I am happy to have the knowledge so kindly shared with me!

I think I'm understanding. I mainly have large architecture in mind, right now, so I think it would not be good for billboards, and I wouldn't want Atlus Textures for in-zone interactive buildings and structures. Although I do have a question about Atlus Textures.

With Atlus textures, am I wrong to think they'd be most efficient for distant scenes, that are not the backgound(mountains, sky, clouds) but they'd always be in sight. Like a distant, giant city with towers that can be seen from anywhere in a zone/area?

Mip-maps: Oh boy! :/ This is opening up a huge world of learning. As I'm definitely not an expert, I'm trying to simplify this in a way that I can understand it, without knowing every nuance, yet. It sounds like mip-mapping may not be very efficient, yet I've read in forums that many MMOs use it.

More Questions.

Are mip-maps only for textures? The LOD is done by the rendering engine, but you can choose between 2 methods, right? One method is to let the engine automatically reduce poly-count, or supply lower-poly versions of your model. From my understanding, you can let the computer do it automatically, or you can manually tell it when to lower count(at what distance) and use a lower-poly model that you made. I'm trying to figure out which may provide a smoother image with the best graphical look. I'm thinking it would be best to go manually. Figure out the good distances to tell the engine when to lower LOD and tell the engine to use the lower-poly model-versions that I make.

Also, are mip-maps only for textures that go on models? or for displaying the texture and the model? For instance: Would I program LOD for the model, then program mip-maps for the textures on those models, to further help/improve performance/improve graphics(if I wanted to use mip-maps)?

I'm going to try and find out what methods Vanguard: Saga of Heroes uses. It definitely looks like they use LOD for many of the structures(buildings, fences, barrels, everything), but I don't know if the artisits designed separate LOD models or if the engine is doing that. I also don't know if they are using Mip-maps or not. Some of the landscapes do seem to use it, but I thought it was the same as LOD, I didn't know mip-maps were a separate function. So I don't truly know the difference, except that I think mip-maps are for textures only and LOD is for models(?)

I also don't know if Vanguard is using billboards or not. I don't think it is. Almost all vegetation and trees move. They all sway, and they actually sway harder, faster, further when it is storming. You can see great distances in Vanguard, due to separating areas into a grid(called chunks) instead of traditional zoning methods. And it is possible to see so far, you can see across multiple chunks and all the graphics are static, not just an illusion while you are in a certain chunk. I can see tiny trees off in the distance and the still sway back and forth. As I approach them, they only appear to get bigger. My guess is that, due to Vanguard's unique chunking method of breaking up game areas, they don't need to use billboarding to save memory.

Well, this is a lot of writing. I will end now. I appreciate your patience and all the help. You've been very helpful and thank you for all the links.

Your thinking is very broad. Just because you think a programme isnt using billboards, could mean it is, but they are implimented very well, and it is the reason why they are used (because at that distance it is hard to tell) BTW, billboards can 'sway', albeit with limited vertices to manlipulate.

LOD is something set within the world editor. Many will use global settings and perform the culling and LOD functions automatically. As most of the actual work is done clientside, if the option is given, to adjust LOD based upon system specs, then the low level poly creation (actually more of a degrading function) is altered to suit. I am not sure what the functions/settings are in MV as I have not tested that far yet. Albeit it is doubtful you would need to carryout so much LOD scritping/functioning yourself, as that is what the world editor is for. Unless you have some specific requirements to do so. If so, then you can create low poly versions of your model to specific settings (there are many good programmes that can do that), script the scene to trigger replacing the models at given circumstances.

Low level poly versions normally are applied with lower versions of the orginal hi-res version of the texture, from the mipmap selection created on either import (into world editor) or clientside rendering. You are right in saying Mipmaps are for textures. LOD (Level Of Detail) is a general term referring to the whole scene, whereby all elements are reduced in resolution, based upon a radius setting from the viewer, in a graduation scale.
Have you ever experienced the 'popping' of terrain meshes as you walk about? This is the LOD portion of the engine adding back polygons/vertices to increase detail.

Atlas textures: As systems become more powerful, so can the clientside rendering functions. Modern systems can support large textures in memory, albeit it is always limited to memory and bus speeds etc. If you have a scene with multiple textures in it, to hold several textures that total say 2048x2048 in overall size, will always take more memory than a single texture of that same overall size. Hence the use of Atlas textures. The drawback is, holding a single texture in memory, but only using a small amount of it. They are only really useful when the scene includes all of its content, like scenery, buildings, npc's that are in view for the majority of the time. Dumping that texture out of memory and replacing it with a new one, when the user changes zones etc.
As you can see from the samples link I gave you, one of the good example uses is terrains, especially those that use overlay texture maps. Mainly because the terrain will always be in view for the most part. You can of course hold several Atlas textures in memory at the sametime. Used correctly, they can save resources.

In an imaginary room, most of the woodwork uses the same set of wood textures, the carpet is fully fitted, the door furniture uses the same metal texture throughout, but some are patterened, the walls use a combination of painted and wallpapered finishes. Collectively you may have 10 seperate textures all loaded in memory at the same time, total resolution is 2048x2048. Each texture has memory registers etc, all taking up further resources. Put all those into a single texture and you can reduce the overheads.
One of the most common uses of Atlas textures is on a character. If you use most modern character creation software, it allows you to change parts of the body, clothing etc. When you commit the model, it will create an atlas texture for that character.