If you are a small studio doing client games, the central problem is not the engine or the gameplay — it is the art pipeline. A two-engineer team can ship a working game loop in six weeks. Making that game look like a game, on the same timeline, is the harder problem.
Over the last 18 months we shipped three back-to-back 3D projects without a dedicated art team. Two were mobile, one was a desktop demo for a venture client. None of them were AAA. All of them shipped on time. Here is the stack we settled on, what we still buy, and the rule we made for ourselves at the start of project three.
The stack, in order of use
Blender. The least surprising thing in this post. We use Blender 4.x for every project — blockouts, level greybox, prop polish, animation, and UV cleanup. The only thing we still bring outside help in for is character rigging on complex humanoids, and even that is shrinking as Rigify gets better.
Mixamo. For humanoid animation cycles — walk, run, idle, basic interactions — Mixamo is still the fastest path from "nothing" to "the character moves." We retarget into our own skeletons in Blender and clean up the obvious foot-sliding. For non-humanoid creatures we animate by hand, because there is no shortcut that has not aged badly.
Quixel Megascans + Polyhaven. Free or low-cost photogrammetry libraries for environment dressing. The trick is to never use a Megascan asset as a hero element. They are great as props, decoration, ground debris. They are not great when they are the thing the player is looking at — they look too real next to anything stylized.
Sketchfab and asset marketplaces. Used surgically. We buy specific assets — a single vehicle, a specific creature — when the project budget supports it. We do not buy "asset packs," because we end up using one model from the pack and the visual language of the other 47 leaks in through implicit decisions.
Substance 3D Painter. Only for hero assets where the texture is doing real narrative work. Otherwise we use Blender's built-in shader graph and accept that not every prop needs to be PBR-perfect.
Where we still hire out
Three places, every time:
- Concept art. One week of contract concept art at the start of a project saves us six weeks of design wandering. We pay for this even on tight budgets.
- Music and SFX. Royalty-free libraries get us through prototyping. For shipping, we contract a composer for the title and one or two cues, and source SFX from a paid library. AI sound generation has gotten interesting, but the legal status is messy enough that we do not ship with it for client work.
- Character art for hero characters. If the game has a main character with a face, we hire a character artist for that one model. Trying to do this in-house with two engineers is the most common way 3D projects start looking "off" without you being able to say why.
The rule we made for ourselves on project three
By project three, we had a recurring failure mode. We would estimate eight weeks of dev, six of those weeks would be code, and the last two would be a panic about visual polish. Then we would slip by three weeks, and the slip would always be on art.
So we wrote a rule:
If we cannot make a level look acceptable using only blockout shapes, free assets, and one paid hero element — the level scope is too large for our team.
That rule has killed at least four levels before they started, and saved every project we have applied it to since. It is not glamorous. It does not feel like "we found the perfect art pipeline." It feels like "we admitted what we cannot do and scoped around it."
Which is, in our experience, how small studios actually ship.
What we are watching
Two things on the horizon, both of which we have piloted but not yet trusted on a paid project:
- Procedural / generative tooling inside Blender. Geometry Nodes is mature enough to do real environment work now. We have done one internal demo with it and the result was promising but slow to iterate.
- AI-assisted texture and material work. We are using it for ideation and for occasional roughness/normal cleanup. We do not ship AI-generated assets in client work without a written conversation about licensing first. That conversation is still surprising people.
If you are starting a small-team 3D project and you want a second opinion on what to outsource and what to keep in-house, we are happy to talk for thirty minutes for free. The reason we offer this is selfish: every conversation we have with someone in the trenches makes our own pipeline a little better.
— K.