How do I find a balance between time and quality when creating game assets for an indie game?

dundyne

I'm making assets for a first person simulator game with semi-realistic graphics, the assets are mostly furniture and decorations. Making both high poly and low poly versions of each asset drastically increases the amount of work needed. Furniture often just consists of simple shapes and it seems unnecessary to bake normals when I can just bevel edges and add weighted normals, then add details in substance. Is this a valid workflow or does it have any downsides I'm not aware of? The alternative would be to create a detailed lowpoly with good topology, then duplicate it and quickly add subd modifier and some edge loops for a high poly bake, but it's rarely as straight forward as that

Any workflows you'd recommend that prioritizes speed but still has decent quality? Polycount isn't a big problem as the game will be small.

Sorry if this has been asked before, but google is basically useless and the forum search is janky

Alex_J

the best use of time is to pay a little money for somebody else's time. i.e. buy some models somebody else made. if you have to do a little work to make everything consistent with your art style that is still usually a lot faster than making everything from scratch

avoiding sculpting and baking will save some time. how much detail and realism you need just depends on your overall art style. choice of your art style is going to be the biggest single factor in how long it takes to make stuff. if you want your couches to be photorealistic that means you need to sit there and work on a single couch for at least a few days. Multiply that time by how many hundreds of random props that you need and your project timeline becomes a year or more with 3/4 of the time making stuff like that. 

i wish i could just do stylized stuff with flat colors but it just doesn't suit me. So to make realism work in reasonable time schedule i pretty much rely on getting high quality base assets and then all my artistic effort just goes into tweaking them. Making things on my own from scratch is only for things I absolutely can't get from elsewhere and absolutely have to have. 

in summary, if the art style is realism i dont think there is a lot you can do to majorly shave time down just with workflow hacks. no matter what, making a realistic couch or trashcan just involves a lot of careful attention that you can't get around unless you reduce quality or replace realism with some form of stylization.

dundyne

@Alex_Jw Hey! Thanks for taking the time to answer. We're just going for a semi-realistic vibe, not archviz quality.  This asset pack is pretty close to what we're aiming for https://assetstore.unity.com/packages/3d/props/grocery-store-furniture-pack-vol-1-238689 

Any advice for a workflow that could achieve these results in a reasonable time?

Fabi_G

Hi! I would say for non-hero assets just use lowpoly meshes with good shading (hard edges at steep angles, bevels with face weighted normals) combined with generic trim atlas and tiling textures. Additional variation can be achieved using unique masks, either vertex painted or texture. Yet another way to create variations would be to add a decal layer on top, wear/damage/grafity/stickers. Look for games similar to yours and learn how they do things.

Alex_J

i'd expect to be able to make 10+ assets from start to finish (in game with textures) in a day at a quality like that would be my speed target. i just use typical box modeling techniques, unwrap and texture. Not sure if using normal maps is going to make a big visual difference for a lot of those things. If you are able to purchase that pack, or a similar one, reviewing the models will tell you a lot about how they are made. 

If you are way off from that speed target I think would just have to look at your modeling techniques to figure out where you can speed it up. Just lots of little things you can do to save time. Moment to moment decision making that just comes from lots of practice. But yeah I wouldnt expect to need to doing much baking at all. Just build the game mesh, texture and that's it. 
if you record video of making an asset it will be easy to point out better ways you might do something. That's not workflow level, just stuff like "don't drag that thing around like a moron, use vertex snapping" or similar.

pior

Considering that you are already aware that the high to low process is a drag ending up in unreasonnable time per asset, one thing you could do is to take one design representative of a final asset (not just a "test"), and run it through all the various processes you know of : high to low, dense mediumpoly, hard edges + rounded shader, and even fully 2000s-era handmade, time logging everything, and taking the asset to a final state. Then compare the results and the time logs. And from there, perhaps establish an tight time target per asset, using the most appropriate technique.

As far as I am concerned I would suggest doing everything as lowpoly with hard edges first, as this allows to :
- put a rough/wip asset in game on day one, and test it in contet
- then you can do base flat colors with temp UVs, and test that in context
- then you can do final UVs and an AO bake, and test that in context
- then you can do final texturing, and test that in context.

And since all the above steps are separate then can be lengthened/shortened depending on schedule and priorities.
Another approach would be to pick a workflow based on how little friction it has. For instance, favoring an approach that lets you do everything in one single software.

Benjammin

To add to the advice above: Look at a set of assets and see what you can share between them. Your grocery store example has a lot of the same shelves and trim pieces that could be reused.

dundyne

@Alex_J I'm able to make maybe 4-5 assets per day if I have references ready to go, I'm mostly struggling with how to finish an asset so it looks smooth before texturing. I've tried different methods and right now I just model all the parts, mark seams, vector paint mask areas that are gonna share materials, join them, turn on auto smooth and add weighted normals, then I manually go through and bevel edges until they look smooth before I triangulate and UV unwrap for export to substance painter. This seems to give pretty decent results, but cylinders and curves tend to still have visible edges from certain angles but that might just be me being too greedy on the amount of segments I add. Would love to hear if I'm doing something wrong

dundyne

@pior Hi, thanks for the advice! I've tried different methods and right now I just model all the parts, mark seams, vector paint mask areas that are gonna share materials, join them, turn on auto smooth and add weighted normals, then I manually go through and bevel edges until they look smooth before I triangulate and UV unwrap for export to substance painter. This seems to give pretty decent results, but cylinders and curves tend to still have visible edges from certain angles but that might just be me being too greedy on the amount of segments I add. Would love to hear if I'm doing something wrong

pior

I'd say the main question is how much time you think is reasonable for an asset to take. If your current approach fits within that and allows you to deliver your project on time, then there's not much reason to stress over it.

And, going beyond all the techniques possible for individual asset creation, there's also the possibility of making all assets from the game use a single material and texture set. Working that way decreases asset creation time tenfold. Not to mention that working on things this way also happens to be extremely satisfying.

Lastly, there's also the question of whether the choice of going for "semi realistic with Substance Painter" is set in stone or not.

On a side note : you haven't shown any representative example of what you are doing (outside of the inspiration/reference) ...

dundyne

@pior We're still in the planning stage creating a list of items we need, while also testing out blockouts ingame for the actual functional pieces. Been messing around with different workflows and styles to prepare for the eventual grind. I've definitely considered using a texture atlas but only have experience with substance painter. I should probably look deeper into that, is there any more to it than to just photoshop a bunch of textures together and align UVs?

gnoop

It depends only on how much risk you can bare.   If your work is not perfect you are replaceable .  If it takes too much time  you are  too expensive.     It's never a simple question.    No risk no successes .  You will never do anything  but low paid job.      Too much self confidence without actually perfect  result  and it's silly looking and useless.      But without confidence in your own vision you will go nowhere either.         Sort of  questions nobody can answer. 

Neox

when i read semi realistic i always kinda shudder, because it is usually not realistic or stylized enough to have appeal, to me personally.

but it could of course mean stylized shapes but more realistic on the materials, or lighting side of things

i think the key is in finding a style that profits from certain workflows and doesnt get better by using others. if you can produce your final art directly, no need for highpoly, retopo, ditch baking, maybe even texturing, if not entirely by using a solid material library you can save a tonne of time and still make something appealing. i think the important thing here will really be defining the artdirection early and defining every production step from there.

when I read furniture, i always think Sims, a great inspiration here is the ad campaign by airbnb, to me this is perfect. most is carried by shading and lighting, i doubt models and textures are complex, the style is super appealing (to me), if i would do a game like that, this is where i would take a lot of notes.