--- name: video-storytelling description: Structure a video's narrative arc and pacing, scaled to its actual duration -- from a short explainer's single arc to a true long-form (8+ minute) piece's chapters, retention curve, and pattern interrupts. Use once the brief's message and audience are confirmed and before writing a shot-by-shot storyboard. Read alongside video-cinematography, which owns per-shot camera/lighting direction once a beat is chosen. --- # Narrative structure and pacing This skill owns the shape of the whole video -- what happens when, and why a viewer keeps watching. `video-cinematography` owns what one shot looks like once a beat is decided. Get the structure right first; a beautifully shot sequence in the wrong place in the arc still loses viewers. The guidance here scales with duration -- rules for a 2-minute explainer and a 12-minute video differ in scope but not in kind. Apply the arc template for short pieces, add chapters and retention management for anything genuinely long-form. ## The explainer arc For a single-arc piece (roughly under 5 minutes), structure the narrative in this order, scaling section lengths proportionally to total duration: 1. **Hook** -- a pattern interrupt or counterintuitive claim, one to two sentences. The opening visual must create curiosity, not just look nice. 2. **Tension / information gap** -- state what the viewer probably believes and why it's incomplete or wrong. Establish why this matters to them. 3. **Concept beats** -- one idea per beat, each building on the last. End each beat with a "but" or "therefore" transition into the next, never "and then" (see "But-therefore, not and-then" below). 4. **Palate cleanser** -- a brief pause, a visual beat, or a "let that sink in" moment between concept beats, giving the viewer a moment to consolidate before the next idea lands. 5. **Key insight** -- the "aha" moment the whole piece has been building to. Follow it with one to three seconds of deliberate silence -- let it land before moving on. 6. **Proof / example** -- a concrete demonstration of the insight working in a specific case, not just an assertion that it's true. 7. **Implications** -- connect back to the viewer: "this means that..." 8. **Reframe and close** -- callback to the hook, restate the core insight in one sentence. ## Hook types Pick one deliberately rather than defaulting to the same shape every time: - **Contrarian**: "Everything you've been told about X is wrong." Best for myth-busting or correcting a common misconception. - **Outcome**: "By the end of this, you'll understand X." Best for concept-driven, systematic explanations. - **Mystery**: "In [year], something unexpected happened..." Best for story-driven pieces with a real narrative to unfold. - **Stakes**: "This one mistake costs people X." Best for practical, how-to-shaped content. ## The 30-second rule The large majority of viewer drop-off on typical platforms happens in the first 30 seconds. The hook and the tension/stakes setup must be fully landed by then -- do not spend the opening 30 seconds on a slow build. A piece that survives this window at a healthy retention rate tends to hold a substantially higher fraction of viewers through the rest of its length than one that doesn't. ## But-therefore, not and-then Never connect narrative beats with "and then" -- it reads as a list, not a story. Connect with "but" (something complicates or contradicts what came before) or "therefore" (something follows as a consequence): > Bad: "Atoms have electrons, and then those electrons have energy levels, > and then..." > > Good: "Atoms have electrons, **but** they don't behave like tiny planets, > **therefore** we need a different model -- **but** that model creates a > new puzzle, **therefore** the actual answer is..." ## Misconception-first Presenting what the audience probably already (incorrectly) believes before revealing the correct picture produces measurably better comprehension and engagement than presenting the correct information cold -- this is a well-replicated finding in educational-video research (Derek Muller's PhD work on physics-misconception videos is the foundational study). For any explainer where the audience likely holds an intuitive-but-wrong model, consider opening with that wrong model before correcting it, rather than starting from the correct answer. ## Guided discovery Rather than stating the answer, reconstruct the reasoning path so the viewer arrives at it themselves (the method long-form science-explainer channels use to keep dense material engaging): 1. Pose a specific, concrete question. 2. Show the obvious/naive approach -- let it partially work, then break. 3. Introduce one new idea -- the key insight -- and pause on it (one to three seconds of silence after a reveal is more effective than rushing past it). 4. Apply the insight step by step, so each step feels like it follows inevitably from the last. 5. Generalize: "this pattern holds beyond this one example." Reveal visuals progressively, layer by layer, timed to arrive exactly when the narration references them -- never show the full picture before the narration has earned it. ## Pacing rules, grounded in cognitive-science research These come from Mayer's Multimedia Learning research (Cambridge University Press) on how people actually process narrated video, not just convention: - **Segmenting**: at most one genuinely new concept every 30-45 seconds. Cramming concepts together measurably reduces comprehension. - **Signaling**: verbal signposts every 30-45 seconds ("here's where it gets interesting") help the viewer track structure. - **Temporal contiguity**: narration and the visual it describes must be simultaneous. Even a few seconds of offset between them measurably hurts comprehension. - **Coherence**: cut interesting-but-irrelevant material. Content that's engaging but doesn't serve the point measurably *reduces* learning on what actually matters -- entertaining tangents have a real cost, not just an opportunity cost. - **Modality**: prefer narration (spoken) over on-screen text for explanatory content paired with visuals -- spoken words plus pictures outperforms written words plus the same pictures for comprehension. Use on-screen text for what needs to be read exactly (a stat, a quote, a claim), not as a substitute for narration. - **New visual element roughly every 3-5 seconds** keeps attention. This is not a cut every 3-5 seconds -- a held shot satisfies it whenever something in frame changes: subject motion, a camera move, a graphic appearing, a reveal. Cutting that often would over-cut a reflective beat (see `video-editing`'s pacing guidance, which governs cut frequency); what this rule asks for is visual *change*, of which a cut is only one kind. ## Scaling to true long-form (8+ minutes) Everything above still applies -- long-form is the same narrative discipline under more time pressure to sustain attention, not a different discipline. What changes at length: ### Chapter structure Break the piece into 2-4 minute chapters (max 5-6 chapters for a 10-15 minute video -- more fragments the narrative). Add explicit chapter markers/timestamps as metadata for the finished piece. **The arc applies at both levels, and they are not the same arc.** The eight-step explainer arc above runs once across the whole piece: one hook at the top, one key insight the entire video builds toward, one reframe at the end. Each chapter then carries a smaller version of the same shape -- a mini-hook or re-engagement line at its start, one idea, a payoff before its end -- feeding the larger arc rather than restating it. A chapter that delivers its own full "aha" and closes it off leaves the viewer with no reason to continue; a chapter that ends on a payoff *and* an unanswered question does. ### The retention curve has known danger zones - **First 30 seconds**: the largest single drop-off point on typical platforms (see "The 30-second rule" above) -- this matters even more at length, since a viewer who leaves in the first 30 seconds of a 12-minute piece never sees the other 11.5 minutes of work. - **Around the 2-3 minute mark**: a second common drop-off point once the opening curiosity is spent and a real payoff hasn't landed yet. Counter it by delivering the first major payoff before 2:00, and placing a pattern interrupt (a visual style change, a B-roll burst, a music shift) around 1:45-2:00 to bridge across it. - **A secondary drop-off point roughly 55-65% through** a longer piece -- counter with a burst of quicker cuts and resolving an open loop planted earlier. - **Re-hook every 3-4 minutes** after the 2-minute mark with a verbal signpost ("but that's not even the interesting part...", "here's where it gets surprising...") to give the viewer a reason to stay through the next chapter. ### Pattern interrupts Deploy a *major* interrupt (visual style change, section transition, music energy shift, a quick burst of several rapid cuts) every 60-90 seconds, and a *minor* interrupt (a B-roll cut, an on-screen graphic, a direct-address line to camera) every 20-30 seconds within a chapter. A viewer going 15+ seconds without any visual or audio change is a real risk point regardless of how good the content is -- attention needs something to track. ### Open loops Plant a question or a "the interesting part is coming" tease in the first 60 seconds, and hold the answer for a real payoff later -- an unresolved open loop is one of the more reliable ways to carry a viewer through a chapter they might otherwise not need to finish. ### End screen Reserve the last ~20 seconds for a close and call-to-action -- do not place essential content there, since platform end-screen elements typically overlay that region. Do not "say goodbye"; tease what's next instead. ## Where this fits - Use `video-creation` to plan the shot list and own the overall brief. - Use this skill to structure the narrative arc and pacing before writing a shot-by-shot storyboard. - Use `video-model-selection` to choose a model per shot, and `video-cinematography` to turn each beat into camera/lighting/consistency direction, once the arc from this skill is in place. - Use `video-editing` to assemble the results according to the chapter/beat structure planned here, `video-quality` before presenting any version as complete.