Best way to begin creating props?

[VisualizeEdits]

I've kind of went past the learning stage of learning how to create basic props and importing it into the engine (back and forth).

Now I want to make sure I can make good effective props that can be used again, but I want to clear some steps that I should take.

1. Create a super-high detailed model in a 3D application.
2. Create LODs (break it down 2x (so 3 total LODs)) of the model.
3. On the super-high resolution model, create a UV Texture for the model with the highest resolution.
4. Use the high resolution texture for the lower LODs.

The reason I was thinking about using the highest resolution texture, is that later during the development phase, when I create graphic settings for the game, I want people to adjust the graphics level (High, Medium, Low) for performance. So I should use the highest texture and reduce the size of it for each graphic level? I also started to have problems smoothing my model. Can't figure out what exactly I need to smooth on my model.

As of the steps I listed above, is this the steps I need to take? Just want to clarify the path I need to go before beginning. Thanks!

[ÅsmundSchei]

You dont want to import many objects which
has like 50.000 polygons. I was told to
model the parts of the object in seperate meshes
to keep the polycount low.

[VisualizeEdits]

You dont want to import many objects which
has like 50.000 polygons. I was told to
model the parts of the object in seperate meshes
to keep the polycount low.

Okay, that make sense now. I'll keep that noted! Thanks Frequia.

[mariomguy]

My friend and I imported his 6-million-poly Z-Brush sculpt into Unreal Engine just fine. Of course, he had a GTX 690, but it was possible. The new engine does not have any polygon limits on import the way UDK had. As long as your GPU can physically render millions of polygons, you are allowed to put all that power into one model if you'd like.

There's so many different methods for creating game models from high-res. Some involve retopology in an application like 3D Coat. Others simply use Z-Brush decimation/reduction methods to get a lower polygon model (better for irregular, organic props). Some just create a blockout model to import something into Z-Brush, do minor sculpting, and then reuse the block model as the game one. The end result just needs to be a model in your game with low enough polys to run in realtime, and textures/normal maps, however you want to do them.

What I do for textures is I lower the resolution until I start to lose meaningful detail if I go any lower. The engine will automatically mip-map your texture sizes down so if you view the object at a distance, a lower mip level is used. A 1k texture might downconvert to the 512x512 mip, then 256x256, 128x128, and so on and so forth the further you view the object. If you want to use a lower quality texture level, just use a lower mipmap.

I'm not sure what "smoothing" your model means. If you mean smooth silhouettes, you need more polygons for that. If you mean smooth normals, just soften the edges.

[VisualizeEdits]

My friend and I imported his 6-million-poly Z-Brush sculpt into Unreal Engine just fine. Of course, he had a GTX 690, but it was possible. The new engine does not have any polygon limits on import the way UDK had. As long as your GPU can physically render millions of polygons, you are allowed to put all that power into one model if you'd like.

There's so many different methods for creating game models from high-res. Some involve retopology in an application like 3D Coat. Others simply use Z-Brush decimation/reduction methods to get a lower polygon model (better for irregular, organic props). Some just create a blockout model to import something into Z-Brush, do minor sculpting, and then reuse the block model as the game one. The end result just needs to be a model in your game with low enough polys to run in realtime, and textures/normal maps, however you want to do them.

What I do for textures is I lower the resolution until I start to lose meaningful detail if I go any lower. The engine will automatically mip-map your texture sizes down so if you view the object at a distance, a lower mip level is used. A 1k texture might downconvert to the 512x512 mip, then 256x256, 128x128, and so on and so forth the further you view the object. If you want to use a lower quality texture level, just use a lower mipmap.

I'm not sure what "smoothing" your model means. If you mean smooth silhouettes, you need more polygons for that. If you mean smooth normals, just soften the edges.

Exactly what I wanted to know! I'm using 32gb ram with 780ti and it's true that there's no polygon limits?? I mean, it make things so much easier (Still keep it reasonable though) to deal with. And by smoothing, I'm not sure. Almost everytime I import a model into Unreal, it keeps telling me that no smoothing group found in the model and don't know what that mean. I'm using the phong (smoothing) tag on my models though.

[mariomguy]

That's just smooth normals. Just soften the edges of your model before export, and make sure you have your model set to export smooth normals, and you'll be all good.

32 GB RAM, and a 780 ti?! Lucky. Yeah, you should be able to get a few million polygons. Once you get into higher numbers, though, the memory of the asset actually becomes a bigger issue than the ability to render polygons. Some massive polygon models can enter the GBs in size, at that point you're going to struggle with bandwidth. But it's still a good practice to limit polygons because A: your game is going to render more than just the one model, B: increasing polygons will limit your efficiency with vertex shaders, and C: not everyone is going to have a 780 ti. I'm still running a wimpy GT 640.

[Laphicet Crowe]

You dont want to import many objects which
has like 50.000 polygons. I was told to
model the parts of the object in seperate meshes
to keep the polycount low.

Explains why when I dissected the assets of a few Unity games, the meshes were in pieces.