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-Frikind longformweek W42date 2026-10-23campaign longform-youtubepillar staff_trainingbeat asset videoduration 156.3sground whitescenes 7

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-W42-Fri 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 7 scenes · 156.3s · comp_id + rendered still + tier + the script

#Layer (comp_id · still · tier)BeatTimecodeMotionLogoAudioVO / on-screen / caption
1s1
matches intent
shared field
signature-3d
open0–24.5sspatial-parallaxicon·3d-extrude♪ bed_in
Today is interviews and discussion, the format that builds trust because real people say real things. The week is about proving you are real, and a good conversation is hard to fake. You will learn how to record an interview and turn one sit down into a stack of clips, a transcript, and a blog post, all from the same recording with a model doing the sorting.
on-screen: Interviews that turn into ten clips
expected on screen: white ground · knot hero in the shared Signal Field · Magister leads · node-graph · spatial-parallax · icon·3d-extrude logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual node-graphshape knotground whitetreatment 3d-extrudemotion spatial-parallaxpower summoninstrument summon→Interviews that turn into ten clips
2s2
matches intent
NumberedList
template
teach24.5–46.1skinetic-buildicon·wireframe♪ node_lock
First step: transcribe before you edit video. We run the recording through Whisper, which writes down every word with timestamps. Now you edit the conversation as text, which is far faster than scrubbing a timeline. Cut a boring answer in the transcript, and the matching video cut comes for free because the words carry the time codes with them.
on-screen: Transcribe first, edit the text
expected on screen: white ground · a NumberedList panel over a dimmed Signal Field · Magister leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id NumberedListvisual node-graphshape knotground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Transcribe first, edit the textcurate items, nodes
3s3
matches intent
ChecklistCard
template
teach46.1–69.9skinetic-buildicon·wireframe♪ node_lock
Second step: ask the model to pull the strong moments. We hand it the transcript and ask for the sharpest, most quotable answers, with the exact timestamps. It comes back with a short list of clip candidates. You make the final call on which ones are real and which are filler. The model reads the whole hour so you only have to judge the highlights.
on-screen: Let the model find the good parts
expected on screen: white ground · a ChecklistCard panel over a dimmed Signal Field · Magister leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id ChecklistCardvisual node-graphshape knotground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Let the model find the good partscurate items, nodes
4s4
matches intent
BuildLog
template
teach69.9–90.4skinetic-buildicon·wireframe♪ node_lock
Third step: auto switch the angles. If you record two cameras, the model can read the transcript and switch to whoever is speaking, building the back and forth edit for you. Remotion lays the switches onto the timeline. You get the polished two angle look without sitting there cutting every turn of the conversation by hand.
on-screen: Two cameras, switched by who talks
expected on screen: white ground · a BuildLog panel over a dimmed Signal Field · Magister leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id BuildLogvisual node-graphshape knotground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→Two cameras, switched by who talkscurate lines, nodes
5s5
matches intent
DiffCard
template
teach90.4–112.0skinetic-buildicon·wireframe♪ node_lock
Fourth step: spin the transcript into writing. We ask the model to turn the cleaned transcript into a plain readable article, real sentences, your guest's actual points, no padding. Now the same hour is a video, a set of clips, and a blog post that people and search engines can read. One recording, many honest outputs, no extra interview.
on-screen: One talk becomes a written post too
expected on screen: white ground · a DiffCard panel over a dimmed Signal Field · Magister leads · node-graph · kinetic-build · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id DiffCardvisual node-graphshape knotground whitetreatment wireframemotion kinetic-buildpower morphinstrument morph+laser→One talk becomes a written post toocurate lines, nodes
6s6
matches intent
KpiGrid
template
proof112.0–134.0sreceipts-counticon·wireframe♪ node_lock
Our conversations go through this exact loop, transcribe, find the moments, switch the angles, write the post. We are not claiming the model decides what is interesting better than you do. It cannot. It just reads fast and lays out the options. The judgment of what is worth saying stays human, which is the only way an interview stays trustworthy.
on-screen: Our own interviews run this way
expected on screen: white ground · a KpiGrid panel over a dimmed Signal Field · Magister leads · receipts · receipts-count · icon·wireframe logo · caption bottom-left
spec (the prompt): comp_id KpiGridvisual receiptsshape knotground whitetreatment wireframemotion receipts-countpower receiptsinstrument spotlight→Our own interviews run this waycurate kpis, statsLabels
7s7
matches intent
shared field
signature-3d
resolve134.0–156.3scoalescenceicon·3d-extrude♪ bed_out
So that is interviews that earn their keep: transcribe first, let the model surface the moments, auto switch the cameras, and turn the talk into an article. One conversation, many real outputs. You can watch our interviews and read the posts at office.temerarii.xyz. The next step is easy, record one ten minute chat and transcribe it before you touch the video.
on-screen: Watch a talk at office.temerarii.xyz
expected on screen: white ground · knot hero in the shared Signal Field · Magister leads · coalescence · coalescence · icon·3d-extrude logo · caption bottom-left
spec (the prompt): comp_id shared Signal Fieldvisual coalescenceshape knotground whitetreatment 3d-extrudemotion coalescencepower coalescenceinstrument coalescence→Watch a talk 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 turn one interview into clips, a transcript, and a blog post Interviews build trust because real people say real things, and a good conversation is hard to fake. This video shows how to record an interview and turn one sit down into a stack of clips, a transcript, and a blog post, all from the same recording, with a model doing the sorting. The method: Transcribe before you edit video. Run the recording through Whisper, which writes down every word with timestamps. Now you edit the conversation as text, far faster than scrubbing a timeline. Cut a boring answer in the transcript, and the matching video cut comes for free because the words carry the time codes. Let the model find the good parts. Hand it the transcript and ask for the sharpest, most quotable answers with exact timestamps. It comes back with a short list of clip candidates, and you make the final call on which are real and which are filler. The model reads the whole hour so you only judge the highlights. Switch the cameras automatically. If you record two cameras, the model reads the transcript and switches to whoever is speaking, building the back and forth edit. Remotion lays the switches onto the timeline. You get the polished two angle look without cutting every turn by hand. Spin the transcript into writing. Ask the model to turn the cleaned transcript into a plain readable article, real sentences, your guest's actual points, no padding. Now one hour is a video, a set of clips, and a blog post people and search engines can read. The judgment of what is worth saying stays human, which is the only way an interview stays trustworthy. Watch our talks and read the posts at office.temerarii.xyz. Keywords: interview production, Whisper transcription, video editing, repurposing content, Remotion, blog from video.

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