Workflow
Where the hours actually go in a 3D project
Efficiency advice for 3D artists usually means shortcuts and hardware. Both help a bit. The bigger wins are in three places most people never measure — and one of them is almost always the largest single sink in the week.
The premise worth getting straight first: efficiency is not speed. Working faster at the wrong thing produces more of the wrong thing. What follows is about removing work that was never contributing to the image — which is a different exercise from doing the same work quicker.
Measure before you optimise
Nearly everyone is wrong about where their time goes, and consistently wrong in the same direction: we remember the interesting work and forget the friction around it. Modelling and lighting feel like the job, so they feel like where the hours went.
For one week, keep a rough tally. Not a rigorous time sheet — just a note each time you switch activity. Almost everyone who does this finds the same surprise: a large fraction of the week went to looking for things, rebuilding things, waiting, and redoing work that was pointed at the wrong target.
The five sinks
01Re-finding things you already own
You have somewhere between 200 GB and several terabytes of models, textures and HDRIs. You cannot see most of it, so checking whether you own a good rusted metal is slower than downloading a new one — and you download a new one. Then the library grows, browsing gets slower, and the loop tightens.
FixMake looking faster than fetching. That is the whole fix, and it is measurable: time how long it takes to get an asset you are sure you own into an open scene. Under thirty seconds and the problem is solved.
02Rebuilding setups you have built before
The same PBR wiring, the same three-point rig, the same render settings, the same output paths. Each instance is ten minutes that feels like one, and it recurs for your entire career. Material wiring is the worst of them because it also fails silently when you get a colour space wrong.
FixSeparate the creative from the mechanical. Authoring a look is creative and worth your time; connecting base colour to base colour is not. Save the first, automate the second.
03Restarting because the brief moved
The single most expensive category, and the least technical. Work gets thrown away because the reference was thin, the brief was verbal, or nobody agreed what "done" meant. Hours of good work deleted because it was pointed at the wrong target.
FixFront-load the agreement. A reference board and a blocking pass shown early costs an hour and routinely saves a day. Get the disagreement to happen when the scene is grey boxes.
04Iterating at full quality
Look-dev at final samples, simulating at final resolution, rendering full frames to check a shadow. Every iteration costs minutes, so you do fewer of them, so each one carries more uncertainty.
FixIterate at the lowest fidelity that still answers the question you are asking. Proxy geometry, low-res HDRIs, region renders, cached sims at low substeps. Fidelity is for the answer, not the question.
05Context switching between applications
Alt-tabbing to a file browser to find a texture, to a web page to download one, to a folder to unzip it, back to the DCC to import it. Each hop is small and each one costs the thread you were holding.
FixReduce the number of windows a routine task touches. Anything that collapses find-download-unzip-import into one action pays for itself many times a day.
The creative / mechanical split
The most useful lens for deciding what to automate. Every task in a 3D project falls into one of two categories, and they deserve opposite treatment.
| Creative | Mechanical | |
|---|---|---|
| Looks like | Deciding the light, the composition, the proportion, the look of a surface | Connecting base colour to base colour, renaming, exporting, setting colour spaces |
| Outcome depends on | Your judgement | The inputs only |
| Doing it twice gives | A different, possibly better result | An identical result |
| Treatment | Protect the time. This is the work. | Template it, script it, or hand it to a tool. |
The test is the third row. If doing a task twice produces the same result both times, your judgement was never involved, and every minute spent on it is pure overhead. Wiring a PBR material is the clearest example: given the same files, the correct graph is fully determined. There is no version of it that is more you.
Mechanical work does not only cost minutes, it costs attention. Every interruption to wire a material or hunt for a file pulls you out of the state where the creative decisions get made well. Removing mechanical work buys back concentration as well as hours, and the concentration is worth more.
Eight habits worth the setup cost
| Habit | Why it pays |
|---|---|
| Build a startup scene | Render settings, colour management, camera, basic rig, output paths. Saves 5–10 minutes per project and eliminates a whole class of delivery mistakes. |
| Name things as you make them | Renaming 200 objects at the end takes an hour. Naming each as you create it takes no time at all and makes selection, scripting and handover possible. |
| Block out before detailing | Grey boxes at correct proportions, shown early. Cheap to change, and it moves the argument to the moment when changing is cheap. |
| Keep a 2K copy of every HDRI | Light with the small one, swap to the large one for the final frame. Scene load times drop noticeably. |
| Use proxies for heavy geometry | Work with stand-ins and swap at render time. Viewport responsiveness is a compounding multiplier on every other task. |
| Save incrementally, always | v01, v02, v03. Zero-padded. Cheap insurance and it makes "go back to the version before I broke it" a ten-second operation. |
| Learn ten shortcuts a month | Not a hundred. Ten, practised until automatic, then ten more. The compounding is real and the cost is nearly zero. |
| Script anything you do twenty times | Batch rename, batch export, repeated setup. The useful scripts are short and you do not need to become a TD to write them. |
Iterate at the lowest useful fidelity
A principle that applies across every stage and is consistently underused. At any moment you are asking the scene a specific question. Match the fidelity to the question, not to the final output.
- "Is the composition right?" Grey boxes and a default light. Seconds per iteration.
- "Is the lighting direction right?" A 1K HDRI and low samples. The broad distribution is all you are judging.
- "Does this material read correctly?" A preview sphere in a neutral environment, not the full scene.
- "Does the sim behave?" Low substeps, low resolution, short frame range.
- "Is the noise acceptable?" A render region on the noisiest area, not the whole frame.
The trap is drifting into final quality early because it looks nicer, which slows every subsequent iteration and therefore reduces how many you do. Iteration count is usually a better predictor of quality than time spent.
Reuse is the compounding one
Everything above saves time linearly. Reuse compounds, because the library you build this year serves every project after it — and an artist five years in with a working library is doing genuinely different work from one with five years of assets they cannot find.
That requires two things. A structure that stays navigable, which is the subject of how to build an asset library that lasts. And visibility, which is the harder half, because Windows shows blank icons for exactly the formats a 3D library is made of — the reason why, and four ways around it.
The promotion rule is the habit that makes it work: nothing enters the library because you downloaded it, only because you used it successfully. That keeps the hit rate high enough that browsing stays faster than downloading, which is the loop you are trying to close.
Where Kiosk fits
Kiosk exists because sinks one and two on this page were the largest ones in my own week, and neither had a good answer.
For re-finding things: it indexes the folders you already have, in place, and renders previews for everything — including the FBX, USD and EXR files Explorer will not show. Texture sets collapse into single tiles, so a folder of 800 files reads as 100 materials. The library becomes something you can scan rather than something you have to remember.
For rebuilding setups: with a live session open in Blender, Cinema 4D, Maya, Houdini or Unreal, clicking an asset puts it in the scene finished. A texture set arrives as a wired material for your active renderer with colour spaces already assigned; an HDRI arrives as a configured dome light. The mechanical half of the job — the half where doing it twice gives an identical result — stops being your job.
For context switching: browsing, the Poly Haven library and export all happen in one window connected to the application you are working in, which collapses find-download-unzip-import into a single click.
None of that touches the creative half, which is the point. It is aimed squarely at the work that was never contributing to the image.
Run the thirty-second test first: time how long it takes to get an asset you are sure you own into an open scene. If the number bothers you, that is the sink worth attacking.
Download the free editionQuestions
What actually slows 3D artists down the most?
Not modelling, and not rendering. The three biggest sinks are re-finding and re-downloading assets you already own, rebuilding setups you have built before — materials, lighting rigs, render settings — and restarting work because the brief or the reference was not nailed down first. All three are avoidable, and none of them are about how fast you push vertices.
Does buying a faster computer make me more efficient?
It shortens render and simulation waits, which is real but usually not where your hours go. If you are iterating on look-dev twenty times a day, faster feedback compounds genuinely. If you render once overnight, a faster machine mostly buys you a shorter overnight. Measure where your day actually goes before spending, because the same money often buys more time back elsewhere.
How do I stop rebuilding the same materials?
Two layers. Save the materials you author into a library your DCC can browse — an asset library in Blender, a gallery or HDA in Houdini, a preset in C4D. And automate the mechanical part: turning a downloaded texture set into a correct shader is a deterministic process, not a creative one, so it should not be costing you ten minutes per surface.
Are templates and scene presets worth the setup time?
Almost always, and the break-even is sooner than people expect. A startup scene with your render settings, colour management, a camera, a basic lighting rig and your usual output paths saves five to ten minutes per project and, more importantly, eliminates the class of mistake where you deliver a render at the wrong colour space or resolution.
How much time should I spend on reference?
More than feels comfortable at the start. Time spent assembling reference before modelling is the highest-leverage time in the project, because every decision afterwards gets faster and fewer of them get reversed. The failure mode is starting to model at minute five and discovering at hour six that the proportions were wrong.
Is it worth learning scripting to speed up 3D work?
For anything you do more than about twenty times, yes. Python in Blender, Maya and Houdini is approachable, and the useful scripts are usually small — batch renaming, batch export, setting up a material from a folder, applying the same settings across selected objects. You do not need to become a TD; you need about a week of getting comfortable.