The 7-element skeleton
hook
The opening move, quoted verbatim from the transcript.
| Field | What it holds |
|---|---|
start_sec / end_sec | Where the hook runs (typically the first 10–30 seconds) |
text | The verbatim hook line, max 25 words |
type | promise | contradiction | big-claim | question | cold-open | story-open |
strength | 1–10, a content-based score of how hard the open earns attention |
rehooks
Every point where the video re-earns attention after the open:
| Field | What it holds |
|---|---|
at_sec | When the re-hook fires |
text | The verbatim line |
function | reopen-loop | raise-stakes | pattern-interrupt | new-question |
heros_journey
The story spine in five slots: who (the protagonist), want (what they’re chasing), obstacle (what’s in the way), turn (the moment it breaks), resolution (what’s different at the end — null when the ending is deliberately open).
beats
The timestamped beat sheet. Each beat: start_sec, a label from Hook · Setup · Authority · Proof · Re-hook · Reveal · Counter · Resolution · CTA, and a one-sentence summary. The beat sequence is the skeleton you can lift most directly — same beats, your content.
ctas
Two slots, soft and hard, each with at_sec, the verbatim text, and a type — soft CTAs are comment / save / like / share; hard CTAs are subscribe / link / product / sub-channel. Where the CTAs sit relative to the beats tells you how aggressive the video’s conversion logic is.
phrases
Repeated verbal tics: text, count, and function (branded-callback | emphasis | filler | signature). Signature phrases are part of a channel’s format — worth noticing, never worth reusing verbatim.
vo
Voice-over delivery: tone, pace_wpm, pauses (minimal | moderate | dramatic), is_ai_voice, and a short cadence note. Useful for casting and for briefing a voice artist — or for deciding whether an AI voice is viable in your niche.
Structure match
The extracted skeleton is compared against 68 known storytelling skeletons curated from 71 outlier channels. You getbest_match, a confidence score, why_it_fits, what_doesnt_fit, an alternative_match, and example channels that use the same skeleton — so you can see the pattern working elsewhere before you commit to it.
Bend it
From the app you can pour your own topic into the extracted skeleton. The engine inherits the structure — hook style, re-hook cadence, beat sequence, CTA placement, VO tone — and writes fresh language for your idea. Two hard rules are enforced: the structure carries over, the wording never does.Diagnostic fields
Alongside the skeleton, the analysis keeps a set of diagnostic reads:hook_strength, emotional_arc, pacing, content_patterns, script_style, virality_signals, key_topics, narrative_summary, cta_present. Reading script analysis is the working guide to those fields — what each value means and when to trust it.
What to do with it
- Start with the beat sheet. Count the beats, note the labels and their timing, then write your script into the same slots. That’s the fastest path from teardown to draft.
- Match re-hook cadence, not re-hook lines. If the reference re-hooks every 20 seconds, your script needs a reason to keep watching every 20 seconds — the specific lines must be yours.
- Check the CTA placement before the CTA copy. A soft CTA at the midpoint and a hard CTA at the close is a different machine than a single hard close.
Related layers
Hooks & re-hooks
The attention machinery, in depth.
Reading script analysis
The diagnostic fields, value by value.
Scorecard
How the execution rates for its format.
What it's selling
The format and meaning layer this skeleton serves.