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-W24-Thukind longformweek W24date 2026-06-18campaign longform-youtubepillar strategic_relationsbeat asset videoduration 142.3sground redscenes 6

Checklist the per-video bar — engine/sim

100.0/100
plain languagevo coverageno dead airuniquenesscaption fitcompletenesscleanliness
quantitative quality · weights learn from your reviews (engine.sim.memory review longform-W24-Thu good|bad)
✓ all static checks pass — one-focal/scene · tier-by-beat · one-track caption · colorway · cast+shape correct · no banned/fabricated. (audio + visual tiers verify on the rendered finals — Phase 2)

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 6 scenes · 142.3s · comp_id + rendered still + tier + the script

#Layer (comp_id · still · tier)BeatTimecodeMotionLogoAudioVO / on-screen / caption
1s1
matches intent
shared field
signature-3d
open0–23.8sspatial-parallaxicon·liquid-chrome♪ bed_in
One machine, eight pillars, still rolling. Today is illustration, the drawn look of a brand. The trap with AI images is that every one comes out in a different style, so your brand ends up a mess of mismatched pictures. We are going to teach you to lock one look and reproduce it, so your illustrations all feel like they came from the same hand.
on-screen: Illustration with one locked style
expected on screen: red ground · tetra hero in the shared Signal Field · Nexus leads · node-graph · spatial-parallax · icon·liquid-chrome logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual node-graphshape tetraground redtreatment liquid-chromemotion spatial-parallaxpower summoninstrument summon→Illustration with one locked style
2s2
matches intent
NumberedList
template
teach23.8–46.8skinetic-buildicon·white-knockout♪ node_lock
Step one. Before generating anything, write your illustration style as a short paragraph: flat shapes, thick lines, your three brand colors, no gradients, plain backgrounds. This is your style sheet in plain English. Every image you ask for will start with this exact paragraph, so the model knows the look before it knows the subject. The words come first; the picture obeys them.
on-screen: Write a style sheet in words
expected on screen: red ground · a NumberedList panel over a dimmed Signal Field · Nexus leads · node-graph · kinetic-build · icon·white-knockout logo · caption bottom-left
spec (the prompt): comp_id NumberedListvisual node-graphshape tetraground redtreatment white-knockoutmotion kinetic-buildpower morphinstrument morph+laser→Write a style sheet in wordscurate items, nodes
3s3
matches intent
ChecklistCard
template
teach46.8–71.3skinetic-buildicon·liquid-chrome♪ node_lock
Step two. Run your prompts through an image model on Replicate or fal, services that let you call a model from a single API request, right from the terminal. You paste your style paragraph, add the subject, and get an image back. Because it is an API call, you can script it, repeat it, and keep every setting identical instead of clicking around a website each time.
on-screen: Generate through Replicate or fal
expected on screen: red ground · a ChecklistCard panel over a dimmed Signal Field · Nexus leads · node-graph · kinetic-build · icon·liquid-chrome logo · caption bottom-left
spec (the prompt): comp_id ChecklistCardvisual node-graphshape tetraground redtreatment liquid-chromemotion kinetic-buildpower morphinstrument morph+laser→Generate through Replicate or falcurate items, nodes
4s4
matches intent
CodeWindow
template
teach71.3–95.1skinetic-buildicon·white-knockout♪ node_lock
Step three. The real trick is the seed, a number that controls the randomness. Use the same seed and the same style paragraph, and the model draws in a steady, matching look across different subjects. Change the subject, keep the seed. Write that seed down in your brand file next to your colors. Now your illustration style is a saved setting, not a happy accident.
on-screen: Reuse one seed to stay consistent
expected on screen: red ground · a CodeWindow panel over a dimmed Signal Field · Nexus leads · node-graph · kinetic-build · icon·white-knockout logo · caption bottom-left
spec (the prompt): comp_id CodeWindowvisual node-graphshape tetraground redtreatment white-knockoutmotion kinetic-buildpower morphinstrument morph+laser→Reuse one seed to stay consistentcurate codeLines, nodes, windowTitle
5s5
matches intent
StatScoreboard
template
proof95.1–118.89999999999999sreceipts-counticon·liquid-chrome♪ node_lock
Our proof is the cast of characters across the studio. We built them in one code-native style and kept it locked, so they read as one family no matter the scene. We did not redraw them per video and hope they matched. We wrote the look down, pinned the settings, and reused them, which is why they feel like one brand and not eight strangers.
on-screen: Our characters hold one look
expected on screen: red ground · a StatScoreboard panel over a dimmed Signal Field · Nexus leads · receipts · receipts-count · icon·liquid-chrome logo · caption bottom-left
spec (the prompt): comp_id StatScoreboardvisual receiptsshape tetraground redtreatment liquid-chromemotion receipts-countpower receiptsinstrument spotlight→Our characters hold one lookcurate pillar, stats, statsLabels
6s6
matches intent
shared field
signature-3d
resolve118.9–142.3scoalescenceicon·liquid-chrome♪ bed_out
So consistent illustration is a written style plus a locked seed, not luck. Write the look in words, generate through an API, pin the seed, save it. Your next step: write three sentences describing how your brand should draw, before you generate a single image. You can see our locked cast holding one look across the whole calendar at office dot temerarii dot xyz.
on-screen: Build your style sheet at the office
expected on screen: red ground · tetra hero in the shared Signal Field · Nexus leads · coalescence · coalescence · icon·liquid-chrome logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual coalescenceshape tetraground redtreatment liquid-chromemotion coalescencepower coalescenceinstrument coalescence→Build your style sheet at the office

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)

youtubeLock One Illustration Style and Reproduce It With a Seed The trap with AI images is that every one comes out in a different style, so your brand ends up a mess of mismatched pictures. This video teaches you to lock one look and reproduce it. What you will learn: - Write a style sheet in words: before generating anything, write the look as a short paragraph. Flat shapes, thick lines, your three brand colors, no gradients, plain backgrounds. Every prompt starts with it. - Generate through Replicate or fal: call an image model from a single API request in the terminal, so you can script it, repeat it, and keep every setting identical. - Reuse one seed to stay consistent: the seed is the number that controls the randomness. Same seed plus same style paragraph keeps a steady, matching look across images. Then the proof: our recurring characters hold one look because they are built on a locked style and reused, not re-rolled. You write the recipe once and the pictures obey it. Keywords: AI illustration, image generation, Replicate, fal, seed, style consistency, brand art. See the whole system run in public at temerarii.xyz.

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