Back Office · office.temerarii.xyz
One asset, all the way in — composition, the wireframe + storyboard, the output format stack, and the template, all read from the SAME content-index record. The expected output matches what /media surfaces for this post.
post longform-W41-Thukind longformweek W41date 2026-10-15campaign longform-youtubepillar it_devbeat asset videoduration 159.5sground blackscenes 8

Checklist the per-video bar — engine/sim

96.7/100
plain languagevo coverageno dead airuniquenesscaption fitcompletenesscleanliness
quantitative quality · weights learn from your reviews (engine.sim.memory review longform-W41-Thu good|bad)
⚠ 2 flag(s) — not yet ship-ready: copy_genericgeneric_scene · see docs/strategy/VIDEO-CHECKLIST.md

Composition comp · template family · expected output

composition LongFormChaptersfamily / template LongFormChapters
9:16 Reelpending1:1 Squarepending16:9 Widepending9:16 4Kpending1:1 4Kpending16:9 4KpendingGIF (SMS)pending
render pending — silent master not yet on disk
expected output: 0/7 rendered — same matrix the /media preview surfaces for this asset.

Composition layer × scene 8 scenes · 159.5s · comp_id + rendered still + tier + the script

#Layer (comp_id · still · tier)BeatTimecodeMotionLogoAudioVO / on-screen / caption
1s1
matches intent
shared field
signature-3d
open0–20.2sspatial-parallaxicon·white-knockout♪ bed_in
One system across surfaces. Today, illustration, the part that usually wanders off on its own. A brand falls apart when the drawings look like they came from a stock bin. By the end of this you will know how to make illustration that belongs to your brand, generated with AI but kept on a leash.
on-screen: Drawings that match the brand
expected on screen: black ground · box hero in the shared Signal Field · Faber leads · node-graph · spatial-parallax · icon·white-knockout logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual node-graphshape boxground blacktreatment white-knockoutmotion spatial-parallaxpower summoninstrument summon→Drawings that match the brand
2s2
matches intent
NumberedList
template
teach20.2–42.9skinetic-buildicon·ember-fill♪ node_lock
First, write the look down before you generate anything. Line weight, palette, whether it is flat or shaded, simple or busy. Put it in your brand file in plain words. Now your prompts are not starting from zero each time. Every image request begins by naming that style, so the model is drawing inside your fence instead of wandering the whole field.
on-screen: Write a style rule first
expected on screen: black ground · a NumberedList panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·ember-fill logo · caption bottom-left
spec (the prompt): comp_id NumberedListvisual node-graphshape boxground blacktreatment ember-fillmotion kinetic-buildpower morphinstrument morph+laser→Write a style rule firstcurate items, nodes
3s3
matches intent
ChecklistCard
template
teach42.9–68.1skinetic-buildicon·wireframe♪ node_lock
Second, when you generate art, force it back onto your colors. We generate a base with an image model, then run it through a step that snaps the palette to our exact brand tokens and clears the background. The raw output is never the final. It always gets pulled onto the brand colors, so a generated drawing and a hand-built panel share the same red and the same edges.
on-screen: Generate, then strip the noise
expected on screen: black ground · a ChecklistCard panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id ChecklistCardvisual node-graphshape boxground blacktreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Generate, then strip the noisecurate items, nodes
4s4
matches intent
CodeWindow
template
teach68.1–91.1skinetic-buildicon·white-knockout♪ node_lock
Third, if you use characters or icons across weeks, lock them. We keep a reference for each recurring figure and feed that same reference back in every time, so the character stays the same person and does not mutate week to week. Consistency is not luck with image models. It is feeding the same locked reference back in on purpose, every single render.
on-screen: Keep recurring characters locked
expected on screen: black ground · a CodeWindow panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·white-knockout logo · caption bottom-left
spec (the prompt): comp_id CodeWindowvisual node-graphshape boxground blacktreatment white-knockoutmotion kinetic-buildpower morphinstrument morph+laser→Keep recurring characters lockedcurate codeLines, nodes, windowTitle
5s5
matches intent
ProcessFlow
template
teach91.1–113.1skinetic-buildicon·liquid-chrome♪ node_lock
Fourth, let the system catch slop. We scan illustrations the same way we scan everything, and flag art that drifted off palette or got too busy for the brand. The model redraws what fails. You are not squinting at hundreds of images deciding which ones feel off. The rule decides, the same way every time, and the misses get fixed.
on-screen: Make a checker catch drift
expected on screen: black ground · a ProcessFlow panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·liquid-chrome logo · caption bottom-left
spec (the prompt): comp_id ProcessFlowvisual node-graphshape boxground blacktreatment liquid-chromemotion kinetic-buildpower morphinstrument morph+laser→Make a checker catch driftcurate nodes, steps
6s6
matches intent
FlowSchematic
template
teach113.1–132.9skinetic-buildicon·ember-fill♪ node_lock
Proof, no numbers. We run a small cast of recurring figures across our whole calendar, and they hold their look from week to week because we lock and re-feed the reference. You can watch the same characters show up, on brand, in piece after piece. That repetition is the demonstration that the leash works.
on-screen: Our cast stays on model
expected on screen: black ground · a FlowSchematic panel over a dimmed Signal Field · Faber leads · node-graph · kinetic-build · icon·ember-fill logo · caption bottom-left
spec (the prompt): comp_id FlowSchematicvisual node-graphshape boxground blacktreatment ember-fillmotion kinetic-buildpower morphinstrument morph+laser→Our cast stays on modelcurate nodes, stages
7s7
matches intent
KpiGrid
template
proof132.9–154.5sreceipts-counticon·wireframe♪ node_lock
So illustration on a brand is generation with rules. Write the style down, snap every output onto your tokens, lock your recurring figures, and let a checker catch the drift. You can see our illustrated brand holding together across surfaces at office.temerarii.xyz. That is how The Big T-M draws in one voice, even with a model doing the strokes.
on-screen: See the illustrated brand
expected on screen: black ground · a KpiGrid panel over a dimmed Signal Field · Faber leads · receipts · receipts-count · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id KpiGridvisual receiptsshape boxground blacktreatment wireframemotion receipts-countpower receiptsinstrument spotlight→See the illustrated brandcurate kpis, statsLabels
8s8
first render · fix pending
shared field
signature-3dgeneric_scene
resolve154.5–159.5scoalescenceicon·white-knockout♪ bed_out
Custom apps + dashboards: own the code and the IP
on-screen: Chapter 7
expected on screen: black ground · box hero in the shared Signal Field · Faber leads · coalescence · coalescence · icon·white-knockout logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual coalescenceshape boxground blacktreatment white-knockoutmotion coalescencepower coalescenceinstrument coalescence→Chapter 7

Format stack 1 aspects · same scenes[], re-cropped

16:9
1920×1080
X/Twitter · YouTube · LinkedIn video

Channels 2 destinations

YouTubeBlog

Social captions supplemental published copy · per channel (comp_id level)

youtubeHow to make AI illustration that actually belongs to your brand A brand falls apart when the drawings look like they came from a stock bin. This video shows how to make illustration that belongs to your brand, generated with AI but kept on a leash so it stays on model week after week. The method: Write the style rule first, before you generate anything. Line weight, palette, flat or shaded, simple or busy. Put it in your brand file in plain words. Now every image request begins by naming that style, so the model is drawing inside your fence instead of wandering the whole field. Generate, then strip the noise. Make a base with an image model, then run it through a step that snaps the palette to your exact brand tokens and clears the background. The raw output is never the final. It always gets pulled onto the brand colors, so a generated drawing and a hand built panel share the same red and the same edges. Keep recurring characters locked. Save a reference for each figure and feed that same reference back in every time, so the character stays the same person and does not mutate week to week. Consistency is not luck with image models, it is feeding the same locked reference back in on purpose. Make a checker catch the drift. Scan illustrations and flag art that drifted off palette or got too busy for the brand. The model redraws what fails, so the rule decides instead of your tired eyes at the end of the day. We run a small cast of recurring figures across the whole calendar, and they hold their look because we lock and re feed the reference. See the illustrated brand at office.temerarii.xyz. Keywords: AI illustration, brand illustration, image models, character consistency, design tokens, brand art.

Cross-links every lens is a view on this one record