Help a newbie out with high poly assets?

Pixel_Fire

Hey Everyone,

I am fairly new to 3D stuff. I have been working with modo for 5 months and ZBrush for 3 weeks. I usually make low poly stuff because that's easy to do and I think about optimization a lot while modeling. 

I am tackling my first high poly asset. I am attempting to make a weapon for Fallout 4 (when the GECK is released). Since I have never really attempted a high poly model before, I have NO idea what is a reasonable count for a game like Fallout 4. 


Here are the concepts of the weapon and the mods it will have.

Here is my base weapon in modo. It is pretty low poly. 

Here is my ZBrush file that I would get the normals from.

Here is a file where I was just messing around with the noise maker and trying for better normals.

Now, here come my questions:
1. Since the base model is low poly, would there be a point in increasing the poly count? I would think the normals could handle most of the obvious faceting. 
2. I am trying to put chains on the weapon in modo, but I don't what super obvious edges on the chains dues to keeping the chains low poly. What is a reasonable count for next gen game weapons? I was told on reddit that 20-50k is fine for weapons. That blew my mind because I never worked on anything higher that 2k.
3. When unwrapping the assets, should I unwrap a low poly version of them and then proceed to make them high poly? Or do I just unwrap the high poly?
4. Should the assets be unwrapped before bringing them into ZBrush?

If anyone has any suggestions or things to point out, I would love if you did. I still have a lot to learn. Also, I am new to this site (made an account 1.5 hours ago), so if the format of this post is not accurate or if it's in the wrong section, please let me know.

Thank you

Bek

Hey. You seem to be misunderstanding the purpose of the highpoly model (if I'm reading your post right). The highpoly does not end up in-game, it's used to bake details down onto the lowpoly. You might find it handy to look at some tutorials that cover the process of making a game-ready mesh from start to finish so you've a good overview of the steps involved and how it all comes together. To answer your questions in order:
 

1. This will depend on how the tangent space normal map bake works out or any specific engine considerations (unlikely to be a major factor, especially for a beginner). Don't worry too much about polycount just yet, though those three vertical loops you have running up the blade are unnecessary.
2. This really depends on context, for fallout I'd imagine 5-30k+ is a typical range, but don't worry about specific numbers, more that you're using geometry in an efficient way (ie, not throwing 30k verts on some small chains when the rest of the model's large pieces are super lowpoly). You might find it useful to go on artstation, filter by marmoset viewer, and use the 'layers' view to see how lowpoly models look topology wise.
3. Highpoly's typically do not need UV's. If they need them you'll know. The lowpoly obviously needs uv's because that's how textures are applied to the model's faces.
4. It's unlikely you will need the UV's inside zbrush. They might come in handy for some purposes (Polygroup via UV's comes to mind) but I think if you need them you'll know about it.

JedTheKrampus

For 2, there are some performance considerations depending on your target platform and renderer that change the types of polycount you can reasonably achieve. For desktop and console GPUs using deferred renderers the rasterizers drop off in efficiency when there's more than one triangle in a 4x4 block of pixels on the render target, and you absolutely want to avoid subpixel triangles for performance reasons, since those tend to wreck performance if you have too many. So, basically, for desktop and console you can use as many polygons as you need to make the object look like the real thing it should look like.

For mobile, you usually want to use as few vertices as possible while still capturing the big forms of the silhouette because all of the geometry is usually sorted into buckets before being rendered one bucket at a time. This is a strategy that mostly decreases the amount of bandwidth required, but tends to fall off quickly the more geometry that it needs to render.

For 3, high polys can be useful to have UVs on if you want to paint your highpoly instead of your lowpoly, but that really depends on the application. It can be really useful to have UVs on your highpoly object in Mudbox and to a lesser extent Mari, but in Zbrush and most other programs it is really not very useful (especially because of Zbrush's Polypaint.) Typically in Mudbox you would want to have UVs on your base mesh which could then be used for baking, to paint your highpoly, or to uniquely identify vertices on your basemesh, which is useful for reimporting morphs on your base mesh. Zbrush tends to use the order of vertices for a similar function. UVs on your lowpoly are always needed.