G—P / FMX 430 / Semester / Week 13

Week 13 · Phase 4 — The Final

Production sprint I

No new territory today. The studio becomes yours: a goal set at 2:00, desk-crits while you build, a demo at 5:15. The only new tooling is for documentation — Sequencer and Movie Render Queue — because the portfolio edit gets made during production, not scrambled together the night before finals.

Session Fri Dec 04 · 10:00 am – 1:50 pm · CCB-134 Bring your locked brief, your sprint plan, everything staged to build
By the end of this session you can

2:00 · Brief The sprint contract

A sprint is a promise you make to the room at 2:00 and keep at 5:15. The brief today is short because the work is yours — but the contract has rules, and so does documentation.

How the sprint runs

At 2:00 sharp, your sprint goal goes on the board: one sentence, demoable, no "work on." ("The door sequence runs start to finish in the headset" — yes. "Work on interactions" — no.) I circulate for desk-crits all afternoon; flag me when you're blocked, and show me the running thing, not the plan. At 5:15, everyone demos against their sentence. Shipping less than you hoped is normal; discovering at 5:00 that nothing runs is a process failure we're here to train out of you.

Documentation starts now

Your final deliverables include a 2–3 minute documentation edit, and the raw material for it is your project as it improves — the blockout-to-beauty arc only exists if you capture the blockout. Sequencer is UE's cinematic timeline: cameras, cuts, and animation tracks over your live level. Movie Render Queue renders those sequences offline at quality your real-time frame budget would never allow. Set both up today, render passes weekly, and the edit assembles itself.

State of the Art · Free markerless mocap changes the sprint

Epic's free MetaHuman Animator markerless plugin extracts full-body animation — fingers included — from single-camera phone video. Character performance, which used to require a mocap stage, is now a production-sprint option for your final. Discussion: what does free mocap do to the economics of indie production?

MetaHuman markerless mocap

2:45 · Sprint Build against the board

Studio time, structured. The steps below are the working discipline, not a tutorial — follow them in whatever order your project demands, except step 1, which is already done, and step 8, which is non-negotiable.

  1. Goal on the board

    One sentence, demoable at 5:15. If you can't phrase it, that's your first desk-crit.

  2. Break it into 45-minute tasks

    Anything bigger than 45 minutes is hiding a surprise. Order tasks so the demo path works earliest — polish last, plumbing first.

  3. Commit before every agent session

    The Week 1 rule at production tempo: clean commit before you prompt, diff review after. Under deadline pressure the temptation is to let the agent run long and trust it — deadline pressure is exactly when that costs the most.

  4. Take the desk-crit

    When I come by, show the running build and name your riskiest unknown. Desk-crits exist to reorder your afternoon, not to grade it.

  5. Cut scope, not quality

    If the goal slips at mid-sprint, consult your cut list from Week 12 and drop a feature whole. A smaller finished thing beats a larger broken one — that's also how the final is graded.

  6. Rough in the documentation sequence

    Create a level sequence in Sequencer with three shots of today's state: an establishing move, a detail, and the interaction that is your sprint goal. Cine camera actors, real cuts.

  7. Queue a Movie Render Queue render

    Set up an MRQ job for the sequence at high quality — anti-aliasing samples up, resolution at final. It can render after class; what matters is the job exists and produces frames.

  8. Push and log

    Before demo: commit, push, AI log entry. The repo at 5:15 is the sprint's actual output; the demo is just its proof.

4:00 · AI lab Your agent, your pipeline

Twelve weeks in, the training wheels are off: no assigned exercise, just your production and your judgment about where an agent earns its keep. The one meta-assignment: notice what your best prompt of the day was, because that's the pipeline knowledge your reflection is made of.

Prompt 1 · The prompt you actually need
[Your call — the task from your sprint board where an agent saves the most time. Written as a spec: current state, desired state, how you'll verify.]

Choose deliberately: sweeps, wiring, and boilerplate are agent territory; composition and feel are yours. Log the prompt verbatim and what came back — tonight's best prompt is a candidate for the "faster" column of your final reflection.

Prompt 2 · Sequencer rough-in
In my level, create a level sequence with three cine camera shots documenting [your sprint goal]: a slow establishing move, a close detail shot, and a shot framing the main interaction. Cut between them in the sequence.

Let the agent do the plumbing — cameras, tracks, cuts — then take over the framing yourself. An agent can build a sequence; it cannot compose a shot you'd put in your portfolio. Scrub the result and adjust every camera by hand.

Prompt 3 · Blocker triage
Here is my top blocker: [symptom, what you expected, what happens instead, any log output]. Diagnose the likely cause in this project and propose the smallest fix.

Smallest fix, stated explicitly — under deadline, agents will happily refactor half your project to fix a typo. Verify the fix in the editor, confirm nothing else changed in the diff, and only then commit.

Before you demo

5:15 · Crit / share Demo against the board

Two minutes each: read your 2:00 sentence aloud, then demo it. Shipped, partially shipped, or cut — say which and why, and name tomorrow's first task. The room offers one triage suggestion per project. Same contract next week, with playtesting on top.

Homework — due before Week 14

TaskDeliverable
Run your second sprint goal at home — same contract, self-enforced.Goal sentence and outcome in the repo; commits to show for it.
Produce a packaged build for playtesting — next week strangers play it cold, without you narrating.Build that installs and launches on your target (headset and/or wall project).
Write your playtest script: what you'll ask testers to do and what you need to learn.3–5 task prompts + 3 questions, committed to the repo.

Resources