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-W42-Tuekind longformweek W42date 2026-10-20campaign longform-youtubepillar multimediabeat asset videoduration 170.2sground whitescenes 7

Checklist the per-video bar — engine/sim

98.2/100
plain languagevo coverageno dead airuniquenesscaption fitcompletenesscleanliness
quantitative quality · weights learn from your reviews (engine.sim.memory review longform-W42-Tue good|bad)
⚠ 1 flag(s) — not yet ship-ready: copy_generic · 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 7 scenes · 170.2s · comp_id + rendered still + tier + the script

#Layer (comp_id · still · tier)BeatTimecodeMotionLogoAudioVO / on-screen / caption
1s1
matches intent
shared field
signature-3d
open0–24.5sspatial-parallaxicon·color♪ bed_in
Today is screenwriting, the script under every video. This is where most brand content dies, in long flat paragraphs nobody finishes. The week is about proving you are real, and a real script sounds like a person talking, not a brochure. You will learn how to write a tight script with an AI model as your rough drafter, while you stay the editor who keeps it human.
on-screen: Screenwriting that does not bore
expected on screen: white ground · octa hero in the shared Signal Field · Lumen leads · node-graph · spatial-parallax · icon·color logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual node-graphshape octaground whitetreatment colormotion spatial-parallaxpower summoninstrument summon→Screenwriting that does not bore
2s2
matches intent
NumberedList
template
teach24.5–47.9skinetic-buildicon·wireframe♪ node_lock
First step: outline beats, not sentences. We tell the model the shape first, a hook, then one idea per scene, then a close. We literally give it a list, open, teach, teach, proof, resolve. The model fills each beat with a draft. Because the bones are fixed, the script cannot wander. You are composing the structure and letting the machine pour words into it.
on-screen: Write the beats before the words
expected on screen: white ground · a NumberedList panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id NumberedListvisual node-graphshape octaground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Write the beats before the wordscurate items, nodes
3s3
matches intent
ChecklistCard
template
teach47.9–73.8skinetic-buildicon·wireframe♪ node_lock
Second step: hold the model to one idea per scene. AI loves to cram three thoughts into a breath. So we add a rule to the prompt, each scene teaches exactly one move and then stops. When a draft scene tries to do too much, you split it. A viewer can hold one idea at a time. The script that respects that is the script people actually watch to the end.
on-screen: One idea per scene, no stacking
expected on screen: white ground · a ChecklistCard panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id ChecklistCardvisual node-graphshape octaground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→One idea per scene, no stackingcurate items, nodes
4s4
matches intent
SchematicCard
template
teach73.8–97.9skinetic-buildicon·wireframe♪ node_lock
Third step: the read aloud pass. We feed the draft to the voice API and listen. Anything that sounds like a press release, you can hear it instantly. Those are the lines to cut. We tell the model to use short, concrete words an eleven year old gets. If a sentence needs a second read, it loses. You are tuning by ear, not by rule book.
on-screen: Read it out loud, kill the fake
expected on screen: white ground · a SchematicCard panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id SchematicCardvisual node-graphshape octaground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Read it out loud, kill the fake linescurate nodes
5s5
matches intent
LegacyCard
template
teach97.9–122.0skinetic-buildicon·wireframe♪ node_lock
Fourth step: keep a banned list. There is a stack of tired words every marketer reaches for, the ones that promise everything and mean nothing. We hand the model that list and tell it never to use them, even as an example. It forces the writing back to real verbs and real steps. The script gets plainer, and plainer reads as more honest, because it is.
on-screen: Ban the empty words on purpose
expected on screen: white ground · a LegacyCard panel over a dimmed Signal Field · Lumen leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id LegacyCardvisual node-graphshape octaground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Ban the empty words on purposecurate nodes
6s6
matches intent
StatScoreboard
template
proof122.0–147.2sreceipts-counticon·wireframe♪ node_lock
Every script in this week, including the one you are hearing, came through this exact loop, beats first, one idea per scene, a read aloud cut, and a banned word list. We are not pretending a model writes finished work alone. It writes fast drafts. The taste is still a person. That split, machine speed plus human ear, is the whole reason the scripts do not sound like robots.
on-screen: Every script here ran this way
expected on screen: white ground · a StatScoreboard panel over a dimmed Signal Field · Lumen leads · receipts · receipts-count · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id StatScoreboardvisual receiptsshape octaground whitetreatment wireframemotion receipts-countpower receiptsinstrument spotlight→Every script here ran this waycurate pillar, stats, statsLabels
7s7
matches intent
shared field
signature-3d
resolve147.2–170.2scoalescenceicon·color♪ bed_out
So that is screenwriting that holds attention: outline the beats, one idea per scene, read it aloud to find the fake lines, and ban the empty words. The model drafts, you decide. You can read the actual scripts behind every video at office.temerarii.xyz. The next step is small, take your next video idea and write the beats before you write a single sentence.
on-screen: Watch the scripts at office.temerarii.xyz
expected on screen: white ground · octa hero in the shared Signal Field · Lumen leads · coalescence · coalescence · icon·color logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual coalescenceshape octaground whitetreatment colormotion coalescencepower coalescenceinstrument coalescence→Watch the scripts at office.temerarii.xyz

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 write a tight script with AI as the drafter and you as the editor Most brand content dies in long flat paragraphs nobody finishes. A real script sounds like a person talking, not a brochure. This video shows how to write a tight script with an AI model as your rough drafter, while you stay the editor who keeps it human. The method: Write the beats before the words. Give the model the shape first, a hook, then one idea per scene, then a close. A list like open, teach, teach, proof, resolve. The model fills each beat with a draft. Because the bones are fixed, the script cannot wander. You compose the structure, the machine pours words into it. Hold it to one idea per scene. AI loves to cram three thoughts into a breath. Add a rule to the prompt: each scene teaches exactly one move, then stops. When a draft scene tries to do too much, split it. A viewer can hold one idea at a time. Read it out loud and kill the fake. Feed the draft to a voice API and listen. Anything that sounds like a press release, you can hear it instantly, and those are the lines to cut. Tell the model to use short, concrete words an eleven year old gets. If a sentence needs a second read, it loses. Ban the empty words on purpose. Hand the model a list of the tired words every marketer reaches for and tell it never to use them, even as an example. That forces the writing back to real verbs and real steps. Plainer reads as more honest, because it is. Every script this week, including this one, came through this loop. The model drafts, you decide. Read the actual scripts at office.temerarii.xyz. Keywords: screenwriting, video script, AI writing, content writing, plain language, scriptwriting method.

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