New to baking normal maps for games objects

dm3d

So I just have a few questions regarding workflow for creating a game asset and I was hoping some experts that make assets for games could shine some light for me.

I've been reading online and it seems like workflow is usually create high poly > Retopo to reduce and optimize polys as much as possible > bake high poly to low poly once uv'd

I put together this picture to show off the different types of normal maps I am getting after a bake, and I don't know which one is correct if any of them even are correct... The materials look fine. 

I am also guessing Mid-Low poly would be for more pc/console games and the low poly asset would be for mobile games. Correct?

I attached the settings I am using for baking the normal map. 

I have been browsing the wiki that exhists here, and stuff.. but it just seems like "proper workflow and techniques" are hard to find

Thank you for reading and I appreciate a response. 

SiegePerilous

Seems like you're on the right page with most of these assumptions. Not sure exactly what you're asking so will ramble slightly.

Your specs...

Looks like you are using a software I don't. I usually bake in Topogun or ZBrush, but whatever works for you or your company is fine I'm sure. Moving obj or other files between different software is a very common thing and pretty standardized so you may find you move from Maya to ZBrush to Topogun to Roadkill (or the way more modern equivalents other people use) and back again quite often. Generally one software does one task well so you go there and do that then move to another.

The best specs to use would be what gets you the best result. I would have to look up what some options are if it became important, but usually defaults are fine until you find they aren't. Then a google search ensues. 1024 was pretty default standard years back when I was in school, but just ask whoever is in charge if they want something else. Sometimes a large object might want a 2048 while a small prop 512 or smaller. Maybe you'll be expected to combine the maps into one 1024x1024 space for multiple objects.

Mid poly or Low poly, again just depends and you'd want to ask whoever is testing you or you work for.

A good question to ask might be what poly count to use for various types of object you want to put in your portfolio? If that's what you're working at.

Thanez

The only thing I would correct you on is the optimizing part. It's not about cutting counts as low as absolutely possible. Optimizing is all about removing any geometry that isn't necessary to achieve the look you want at the screen resolution you intend it to be viewed at. Sometimes you'll run into a situation where you can either have a chamfered edge with 4 verts per edge, softened, or to collapse that chamfer into only 2 verts per edge. Yay, optimized! But if you then have to harden that edge because the normal map can't cope, you're back up to 4 verts per edge. At that point you might as well have kept the nice chamfer which would look better up close, and put less strain on the normal map.
I'd like to point out that I like your lowpoly better than your midpoly, because even though you used half the geometry on the low, you used it twice as well. Basically, if you only showed me the "material preview" shots, and asked me to guess which is 130 tris, and which is 60tris, I'd say your lowpoly was 130.

Where you put the splits in the UVs is gonna be different for each model. As every split along an edge doubles the vert count along that edge, I suggest trying to use as few as possible, to the point where your normal map breaks down crying. Then do a UV split there.

A "proper" workflow is one that gives you consistent results based on what kind of model you're making. My current is make lowpoly->UV->duplicate and add supporting geo to get highpoly->bake->texture lowpoly->make custom LODS and bake textures from lowpoly and normals from high.
The reason I UV before generating the high from it is that I then end up with a UV'd highpoly, which is useful for some effects I want. One upside is that when I'm done texturing the low, I can bake the textures onto the high, and I can get some very pretty renders without using tesselation and all the hassles that come with that.

The way you test, show, then ask is the correct way forward because the more effort you put into your question, the more effort we'll put into the answers.
You're on the right path. The only way to fail is to not achieve the look of what you set out to.

Eric Chadwick

Some step by step here, and explanations why, plus all-important caveats

http://wiki.polycount.com/wiki/Texture_Baking

poopipe

The idea is to remove geometry that doesn't contribute to the silhouette.  

Your bake results are fine - or at least look fine at the size I'm viewing the-and you've got sensible bake settings so good stuff on the whole. 

 you could get a closer representation with the low res by moving a few verts to better fit the high res - you can see the mismatch in silhouette in the images you posted fairly clearly and it's be pretty simple to fix. 

dm3d

Wow you guys! This is awesome thank you so much for the information.  Somehow I missed the Texture_baking Wiki... and Ill look through that.