I have been sent a lot of scene files. Students, applicants, people who bought a course and got stuck.
The render is almost never the problem. The file is.
Every article about art tests is about days
80.lv collected eight artists on how they approach tests. Jacob Claussen finished a seven-day test in a weekend, under a 10,000 polygon limit.
Natalia Gutiérrez used all twelve of the days she was given.
Skillsearch argues the other side, twice. Their advice to candidates is to use the full week and never submit early. Their advice to studios is to keep tests short, because long ones make good people walk away.
Both are arguments about time. Neither says what happens in the first minute after you press send.
The answer is buried in the 80.lv piece
Clinton Crumpler is quoted there on what he checks. "Texture attention and clarity and texture density are my first looks." Quality, in his words, comes after that.
That is one sentence inside a long article about days. It is the most useful sentence in it.
80.lv does tell you to hand in clean files and use naming conventions. It says it once, near the end, under presentation. I think it belongs at the top.
So I measured my own machine
I ran an outliner check over every Blender file I have here. 46 files, 831 mesh objects, 15.7 million triangles.
The check looks for what a lead sees before opening a render. Unapplied scale, default names, origins away from the mesh, missing UVs, overlapping UVs, unused materials, texel density jumping between objects.
Fifteen of the 46 passed everything. Thirty-one did not.
The counts
Across those 831 objects:
- 290 still had unapplied scale, 52 of them non-uniform
- 85 had no UV map at all, and 110 had one created and never used
- 224 of the 618 I could test had over one per cent of the UV layout covered twice
- 602 still carried a .001 style suffix
- 506 were still called Cube, Circle, Plane or Cylinder
- 161 materials had no user at all
The suffix problem is bigger than the object count suggests. Counting meshes, materials and images too, 38 of the 46 files had at least one. 2,502 altogether.
Texel density is the one that reads as careless
Thirteen files had three or more unwrapped objects, so I could compare across a set.
Ten of those thirteen varied by more than four times. I measured it as texels per metre on a 1024 map.
A reviewer notices uneven density without always knowing why. The arithmetic behind it is in the 4K article.
The worst file took twenty-one seconds
Somebody sent me a file on Telegram called Test.blend. It holds 384 mesh objects and 1.26 million triangles.
224 of them have unapplied scale. 373 carry a duplicate suffix. 334 are still named after the primitive they started as.
Its texel density spans a factor of four hundred thousand. The whole check took 21.7 seconds.
Three minutes for forty-six files
The full run was 180 seconds of wall clock, and that includes launching Blender 46 times. Blender spent 102.7 seconds on the checking itself.
That is 2.2 seconds a file. A human clicking through an outliner is slower, but not by much.
Sit with that. Nine days of work, sorted by something shorter than the time you spend picking a thumbnail.
My own files failed too
Four of the 46 are product source files I sell. Three passed clean. One was flagged for overlapping UVs.
That one is mirrored on purpose. My check cannot tell the difference, and a lead scrolling your file cannot either.
My scratch files were worse. Untitled2.blend has 136 objects, 107 duplicate suffixes and 81 primitive names. I am not writing this from a clean desk.
What to do with the last hour
Do not spend it on one more render pass. Spend it here:
- Apply scale and rotation on everything.
- Rename every object. Anything ending in .001 is a rename you skipped.
- Put origins where the object is used from, not where Blender left them.
- Compare texel density across the whole set, not one object at a time.
- Delete unused materials and empty material slots.
- Write one line in your email for anything that looks wrong on purpose.
Six is the one people skip. Mirrored UVs, a deliberate density jump, a stack of trim pieces. Say it, or it reads as an accident.
Most of this is decided at unwrap time, which is what the unwrap course is about. Density and bakes are the Painter course. The presentation side I already covered in the portfolio article.
None of this turns weak art into strong art. It stops strong art being read as careless in the first two seconds.