AI for friction, humans for meaning. Raw output is never finished work. Verify before trusting. Disclose where it matters. Skepticism welcome when specific.
Standing recap.
Today
Today's promise
You'll make AI write working Blender Python for you — zero coding knowledge required —
and you'll spot, in ten seconds of wireframe, why "free AI 3D model" usually isn't free.
Both halves happen to every student today, hands-on.
The week in one line
You don't need to learn Python. You need to describe, run, and report.
Board: DESCRIBE → RUN → REPORT → RE-RUN.
Deliverable
You will leave with
A 3D scene plan — lights with motivations, honest budget
One keeper — a script or checklist you'll actually reuse
"The keeper is the real deliverable — the thing that saves you an hour next week."
Runs in the lab on a provided mesh. Studios literally pay for this judgment now.
Before and after the build
Scene plans + render QA
Plan: frame contents · asset sources · every light has a motivation · camera intent · honest budget
Triage: "grainy shadows, Cycles, 40-min frames" → diagnostic order
AI drafts both; the motivations and verdicts are yours.
Demo
The scaffold, the dials, the crash
~1:15. Pre-tested script ready. Prompts from DEMOS file.
Demo · move 1
The ask
Write a Blender 4.x Python script that builds a previz scene: ground plane with subtle displacement, 12 low-poly rocks scattered without overlaps, three-point area lighting, camera framing the cluster from low angle, light volumetric fog. Comment every section in plain language for an artist who doesn't code. List what could go wrong when I run it.
Read the comments aloud when it returns. Read its own risk list.
Demo · move 2
The run
Scripting tab → paste → Run → orbit
One paste. That used to be forty minutes of setup.
Let the scene land. Orbit slowly.
Demo · move 3
The dials
Rocks: 12 → 40. Fog: doubled. Re-run.
A script is a dial, not a black box.
Invite a student to call a value. Change it live.
Demo · move 4
The planned crash
It failed. Good.
Read the traceback. Copy ALL of it. Paste it back.
"You don't need Python. You need copy-paste and honesty about what happened."
Let the error be scary for two seconds, then run the loop. Fix usually lands first try.
Demo · move 5
The artist's judgment
What does this scene still need that no script provides?
Composition intent · story · materials with meaning · light motivation
Ask the room first. Close with the render checklist request — previz vs portfolio still.
Break
Break — 10 minutes
Lab next: your scaffold, your verdicts.
Share-folder link slide staged.
Lab · 70 minutes
Scaffold → Reality Check → Keeper
Class scaffold + 2 values (15) → YOUR scaffold (20) → Reality Check (15) → Plan + keeper (20)
Parts + times on board. Share folder link visible.
Lab · step 1
15 min
Run the class scaffold
Paste the provided script → Run → change two values of your choice → re-run.
You have now edited Python. That's the whole trick.
Guaranteed success first builds courage for step 2.
Lab · step 2
20 min
Your scaffold
Request one for YOUR project: Week 6 character's set · degree-show space · product turntable.
Errors are wanted. Run the error loop. That's the skill.
Celebrate every completed error loop aloud. Check they name their Blender version.
Lab · step 3
15 min
Reality Check
The provided AI mesh → wireframe pass → 5 written findings + a verdict:
background / blocking / retopo (cost it) / reject
Push past "it's messy" to named crimes: density, flow, backside, ngons, scale.
Lab · step 4
20 min
Plan + keeper
Scene plan for your real project — lights with motivations.
The keeper: one script or checklist you'll actually reuse. Save the prompt WITH the script.
"Which of these will you genuinely run next week? That one's the deliverable."
Always available
No Blender?
The planning track
Scene plan + light motivations + script spec + Reality Check on annotated screenshots.
The director's version of the same knowledge. Equal credit.
Or pair as director/operator with a Blender owner.
Share-out · 25 min
Keepers + verdicts
Keepers demoed live — would you actually reuse this?
Reality Check verdicts defended — and why?
One planning-track scene plan read — the light motivations especially
The keeper demos are quick and satisfying — run 3–4.
Every week
Integrity check
This week's four
Read your script for destructive ops before running?
AI meshes: disclosed in a portfolio breakdown?
Can you state why each light exists?
Save the prompt with the script — it's a dependency now
#2: "'Background rocks AI-generated, cleaned; hero modelled by me' is exactly how studios talk."
Reflection
Write 4 lines
The value you changed (and what it taught you about dials)
Your sharpest Reality Check finding
Your keeper — and when you'll next run it
One thing in 3D you'd never delegate
Two quiet minutes.
Next Wednesday
Audio, Voice, Video, and Multimodal Critique
Transcripts, captions, temp VO — and critique workflows that work on any visual work you make.
Bring 60 seconds of any rough footage or a WIP still.
End on time; corridor questions after.
Tonight, optional
Go deeper
Scripting for Artists — Blender's official series (embedded on week page)
Blender Python API docs — know they exist
Blender Studio open projects — what GOOD topology looks like
The gift script: kill a chore for a classmate's workflow