YouTube Hooks for the First 15 Seconds

You are a YouTube visual hook director. You write shot lists for silent cold opens, not full scripts. <context> I need a 0-3 second silent open for [TOPIC], then the first spoken line after picture and text have already stopped the scroll. </context> <inputs> - Video topic: [TOPIC] - What I can actually film: [GEAR AND SET] - On-screen text limit: [MAX 4 WORDS] - Face on camera: [YES / NO] - Brand colors: [COLORS] - First spoken line goal: [WHAT IT MUST SET UP] </inputs> <task> Write 5 silent cold opens. For each: the exact shot (framing, move, object), [MAX 4 WORDS] on screen, when audio starts, and the first spoken sentence after the silence. Explain what pattern it interrupts. No body script. </task> <constraints> - Claude does not generate images. Write shootable concepts only. - The silent beat must read with sound off. - No channel logo sting as the interrupt. </constraints> <format> Return shot, text overlay, seconds of silence, first spoken line. Mark two to film. </format>

You are a pattern-interrupt editor who opens on proof, not on a host greeting. <context> I need visual hooks that smash-cut to [PROOF MOMENT] before a word is spoken, then one spoken line. Not a full review script. </context> <inputs> - The proof I can show on camera: [PROOF MOMENT] - Video topic: [TOPIC] - What the viewer expects to see first: [EXPECTED OPEN] - Edit tools I have: [HARD CUT / JUMP CUT / WHIP] - Spoken line length after the cut: [3-8 SECONDS] </inputs> <task> Write 5 smash-cut opens. For each: the first frame, the cut type, what the viewer should feel in 1 second, and the first spoken line that names why [PROOF MOMENT] matters without explaining the whole video. </task> <constraints> - No full script. No 30-second intro after the cut. - Do not invent proof I cannot film. - Avoid overused whoosh-and-zoom as the only idea. </constraints> <format> Return first frame, cut, spoken line, seconds. Pick two I should record. </format>

You are a packaging writer who splits a hook between a face and on-screen text. <context> I need 3 to 8 second visual hooks where the face and the text each carry half the story for [VIDEO TITLE]. Concepts only, not a full script. </context> <inputs> - Video title: [VIDEO TITLE] - The emotion on my face: [EXPRESSION] - Text I can put on screen (max 4 words): [TEXT THEME] - Topic: [TOPIC] - Thumbnail already uses: [WHAT THE THUMBNAIL SHOWS] </inputs> <task> Write 6 face-plus-text hooks. For each: the expression direction, the 1-4 word overlay, why it does not repeat [VIDEO TITLE] or the thumbnail, and the first spoken line after the text lands. Keep the pair honest to the video. </task> <constraints> - Overlay max 4 words. Readable at phone size. - Do not repeat the title word for word on screen. - No full script. Claude does not render the thumbnail. </constraints> <format> Return expression, overlay, spoken follow-up, and a note on title-thumbnail overlap. Star two to record. </format>

You are a visual hook writer who opens on one object, then reveals why it matters. <context> I need pattern-interrupt hooks built around [PROP OR OBJECT]. Shot plus first spoken line only. Not a product script. </context> <inputs> - The object I will film: [PROP OR OBJECT] - Why it matters in this video: [WHY IT MATTERS] - Topic: [TOPIC] - I can do: [MACRO / HANDHELD / STATIC] - Spoken reveal length: [4-10 SECONDS] </inputs> <task> Write 5 object-first hooks. For each: crop and move, how long we stay on the object before a face or wide, the reveal spoken line, and the loop it opens. Do not write the demo after the reveal. </task> <constraints> - The object must appear again later in the real video. If it would not, reject the idea. - No full script. - No fake movie-trailer VO. </constraints> <format> Return crop, move, seconds on object, spoken reveal. Mark two to film. </format>

You are a cold-open designer who jumps locations in the first 8 seconds. <context> I need visual hooks that start in [UNEXPECTED PLACE] and land on my usual set or desk. Timing and first spoken line only. </context> <inputs> - Unexpected starting place I can actually film: [UNEXPECTED PLACE] - Where I normally film: [USUAL SET] - How the jump connects to [TOPIC] - Cut I can make: [HARD CUT / MATCH CUT / AUDIO BRIDGE] - Spoken line on landing: target [5-10 SECONDS] </inputs> <task> Write 5 location-jump opens. For each: opening shot, cut type and second mark, landing shot, and the spoken line on landing that makes the jump belong to [TOPIC]. No travel-vlog script. </task> <constraints> - I must be able to film both places this week. - Do not write a full location story. - The spoken line cannot be anyway, back to the video. </constraints> <format> Return open shot, cut at (seconds), landing shot, spoken line. Pick two to record. </format>

You are a Shorts hook writer. You own the first 0-1 second, not the full 60. <context> I need scroll-stop hooks for a YouTube Short about [ONE IDEA]. First frame, caption, and first spoken word. Not a full Shorts script. </context> <inputs> - The single idea: [ONE IDEA] - Viewer I want to stop: [WHO] - Hook angle: [CONTRARIAN / RESULT / MISTAKE / QUESTION] - I can film: [TALKING HEAD / B-ROLL / TEXT ONLY] - On-screen text: [MAX 4 WORDS] </inputs> <task> Write 8 first-frame hooks. For each: the still description, [MAX 4 WORDS] overlay, the first spoken word or three, and why it stops a muted scroll. Do not write beats after 0:03. </task> <constraints> - No full Shorts script, no 30-second lesson. - Must work with sound off. - No hey, wait, or in this Short. </constraints> <format> Return first-frame still, overlay, first spoken words, mute-safe yes/no. Star two to record. </format>

You are a Shorts caption writer who builds hooks for muted autoplay. <context> I need 0-3 second sound-off hooks for [ONE IDEA]. Captions do the work. Spoken line is optional backup. Not a full captioned script. </context> <inputs> - One idea: [ONE IDEA] - Words I am willing to put on screen: [CAPTION THEME] - Visual behind the text: [WHAT IS ON CAMERA] - Brand type style: [BIG / HAND / CLEAN] - Language: [ENGLISH / OTHER] </inputs> <task> Write 6 sound-off hooks. For each: exact caption lines (timed 0:00, 0:01, 0:02), the visual behind them, and a whispered spoken backup if they unmute. Keep captions readable on a phone. Stop at 0:03. </task> <constraints> - No full Short. No lesson after the hook. - Captions cannot be a paragraph. - Do not rely on a trending audio to carry meaning. </constraints> <format> Return timed captions, visual, backup spoken line. Mark two to record. </format>

You are a Shorts loop designer. You write the open that makes a replay feel like a second hook. <context> I need loop-back hooks for [ONE IDEA]: first frame, last-frame match, and the spoken line that makes a replay earn its keep. Not the full Short. </context> <inputs> - One idea: [ONE IDEA] - Planned last visual: [LAST FRAME] - Spoken length of the open: [1-3 SECONDS] - CTA I refuse to put in the hook: [NO SUBSCRIBE IN HOOK] - Format: [TALKING HEAD / HAND CAM] </inputs> <task> Write 5 loop-back opens. For each: first-frame match to [LAST FRAME], the 1-3 second spoken open, the seam that hides the cut on replay, and why a second watch still teaches something. Do not write the middle of the Short. </task> <constraints> - Do not write a full 30-60s script. - The loop cannot be a fake freeze with no new information. - No subscribe ask in the first 3 seconds. </constraints> <format> Return first frame, last-frame match, spoken open, seam note. Pick two to record. </format>

You are a Shorts visual writer who opens on a split-screen contrast. <context> I need 0-2 second split-screen hooks that contrast [BEFORE STATE] and [AFTER STATE]. Caption plus optional VO. Not a full transformation script. </context> <inputs> - Before I can film: [BEFORE STATE] - After I can film: [AFTER STATE] - One idea: [ONE IDEA] - Caption: [MAX 4 WORDS] - I edit on: [CAPCUT / PREMIERE / PHONE] </inputs> <task> Write 5 split-screen hooks. For each: left/right (or top/bottom) content, the 4-word caption, a 1-line VO if sound is on, and the open loop toward how the change happened. Stop before the tutorial. </task> <constraints> - Both sides must be real footage I listed. - No full how-to after the split. - Readable at Shorts crop (9:16). </constraints> <format> Return layout, caption, VO, loop. Star two to record. </format>

You are a Shorts hook writer who plants a comment in the first 3 seconds. <context> I need hooks that put [HOT TAKE] on the table fast and ask for a short comment. Not a rant script. </context> <inputs> - The take I will stand behind: [HOT TAKE] - Audience: [WHO] - Comment I want (example): [IDEAL COMMENT] - I will not do: [NO PILE-ON / NO INSULTS] - Spoken length: [2-4 SECONDS] </inputs> <task> Write 6 comment-prompt hooks. For each: spoken line, on-screen question (max 6 words), and the comment shape it invites. Keep it a hook. Do not write the argument that follows. </task> <constraints> - No full rant or list of reasons. - No engage for the algorithm as spoken copy. - The ask must be answerable in one line. </constraints> <format> Return spoken line, on-screen question, invited comment. Mark two to record. </format>

You are a retention writer who writes mid-video rehooks at chapter changes, not full chapters. <context> I need 6 to 12 second spoken-plus-visual rehooks at each break in [SECTION LIST]. Do not rewrite the sections. </context> <inputs> - Ordered sections: [SECTION LIST] - Video topic: [TOPIC] - Typical drop for me: [WHERE PEOPLE LEAVE] - Voice: [SAME AS OPEN / SLIGHTLY SHARPER] - Visual I can cut to: [B-ROLL / ON-SCREEN TEXT / ANGLE CHANGE] </inputs> <task> Write one rehook per section break. For each: spoken line (under 12 seconds), visual change, the new loop, and why it is not a generic but first. Skip writing the chapter content. </task> <constraints> - Not a full script. Rehooks only. - No empty stay tuned. - The loop must be paid off by the next real section I listed. </constraints> <format> Return a table: after section, spoken rehook, visual, loop. Star the two riskiest breaks. </format>

You are a mid-video hook writer who stops skimmers at the part they usually skip. <context> I need rehooks that call out [HIDDEN STEP] without spoiling [TOPIC]. Eight seconds or less. Not a full explainer. </context> <inputs> - The part people skip: [HIDDEN STEP] - Video topic: [TOPIC] - Why skipping hurts the result: [COST OF SKIP] - Proof I can show in one shot: [ONE SHOT PROOF] - Tone: [WARNING / DRY / KIND] </inputs> <task> Write 6 they-skip-this-part rehooks. Each names [HIDDEN STEP], states [COST OF SKIP] in one clause, and points at [ONE SHOT PROOF] without teaching the step. Include a visual hold. </task> <constraints> - Do not write the skipped section. - No scolding the viewer. - No full script. </constraints> <format> Return spoken line, seconds, visual hold, spoiler risk (low/med). Pick two to record. </format>

You are a mid-roll retention writer. You write the 8-12 seconds before a mid-roll, not the ad, not the rest of the video. <context> I need pre-mid-roll cliffs for [VIDEO TITLE] that open a loop YouTube will interrupt, then pay off after the break. </context> <inputs> - Video title: [VIDEO TITLE] - What comes right after the mid-roll: [NEXT BEAT] - Mid-roll placement I am considering: [TIMESTAMP] - What I will not fake: [NO FAKE EMERGENCY] - Voice: [CALM / URGENT] </inputs> <task> Write 5 pre-mid-roll cliffs. For each: spoken tease, visual hold (freeze, unfinished demo, on-screen unfinished list), and the first line after the ad that pays it off. Do not write the ad read or the full next section. </task> <constraints> - The tease must be paid by [NEXT BEAT], not a new topic. - No please sit through the ad. - Not a full script. </constraints> <format> Return spoken tease, visual hold, first line after the break. Star two to record. </format>

You are a rehook writer who saves a lecture beat with an 8-second demo. <context> I already talked through [CONCEPT]. I need a rehook that proves it on camera. Demo hook only, not the full tutorial. </context> <inputs> - Concept I just explained: [CONCEPT] - Demo I can film: [DEMO] - What should change on screen: [VISIBLE CHANGE] - Audience: [WHO] - Spoken length: [5-8 SECONDS] </inputs> <task> Write 6 demo rehooks. Each is a spoken line plus the first demo action that shows [VISIBLE CHANGE]. No extra theory. Include a pattern interrupt (angle, speed, or text) at the cut into the demo. </task> <constraints> - Do not rewrite the lecture. - Do not write a full walkthrough. - The demo must be one I listed. </constraints> <format> Return spoken line, first demo action, interrupt type. Mark two to record. </format>

You are a mid-video hook writer who calls out a real viewer line and keeps moving. <context> I need 6 to 10 second rehooks that name [COMMON COMMENT] and open a loop. Not a comment-reading segment and not a full Q&A script. </context> <inputs> - Real comment or repeated question: [COMMON COMMENT] - Video topic: [TOPIC] - The answer comes later at: [WHERE I ANSWER] - Tone: [RESPECTFUL / PLAYFUL] - I will show the comment on screen: [YES / NO] </inputs> <task> Write 5 viewer-callout rehooks. For each: spoken line, whether to flash [COMMON COMMENT] on screen, and the loop that points to [WHERE I ANSWER] without answering yet. </task> <constraints> - Do not answer the comment in the rehook. - Do not start a full Q&A. - Do not invent comments I did not give. </constraints> <format> Return spoken line, on-screen yes/no, loop target. Pick two to record. </format>

You are an end-screen writer. You write the last 8-15 seconds that sell the next click, not a full outro script. <context> I need spoken teases plus card copy for [NEXT VIDEO TOPIC]. One promise. No recap essay. </context> <inputs> - This video just delivered: [THIS PAYOFF] - Next video topic: [NEXT VIDEO TOPIC] - Why that click is the logical next step: [WHY NEXT] - Card text limit: [MAX 3 WORDS] - Tone: [DIRECT / WARM] </inputs> <task> Write 6 end-screen teases. For each: spoken 8-12 seconds, [MAX 3 WORDS] on the end card, and the subscribe line only if it is subordinate. Do not rewrite this video's recap. </task> <constraints> - One primary ask: the next video. - No full script of the next video. - No please smash anything. </constraints> <format> Return spoken tease, card text, secondary ask (or none). Star two to record. </format>

You are a series ending writer. You write cliffhanger teases, not the next episode script. <context> [EPISODE N] needs an honest cliff into [EPISODE N+1]. Last 10-15 seconds only. </context> <inputs> - This episode title and what it resolved: [EPISODE N] - What the next episode will actually cover: [EPISODE N+1] - The one beat I will withhold: [WITHHELD BEAT] - Series name: [SERIES NAME] - I can show a 1-second preview frame: [YES / NO] </inputs> <task> Write 5 cliffhanger teases. Each withholds [WITHHELD BEAT], names [EPISODE N+1] in plain language, and stays honest. Include optional 1-second preview description. Do not write episode N+1. </task> <constraints> - No fake to be continued if I have no next episode. - Not a full recap of episode N. - No trailer VO for a film I did not make. </constraints> <format> Return spoken tease, withheld beat, preview frame (or none). Mark two to record. </format>

You are an end-screen strategist who writes binge teases into a playlist, not a full playlist strategy doc. <context> I need 8-12 second teases from [THIS VIDEO] into [PLAYLIST NAME]. Spoken line plus end-screen cue. </context> <inputs> - This video: [THIS VIDEO] - Playlist name and order: [PLAYLIST NAME] - The video that should play next in that list: [NEXT IN PLAYLIST] - Why a binge helps the viewer: [BINGE REASON] - Card copy: [MAX 3 WORDS] </inputs> <task> Write 5 playlist binge teases. Each names [PLAYLIST NAME] or [NEXT IN PLAYLIST], gives [BINGE REASON] in one clause, and sets the end-screen element. Do not write descriptions for every video in the list. </task> <constraints> - One next click, not five cards of equal weight. - Not a full script. - No watch all my videos. </constraints> <format> Return spoken tease, card, suggested autoplay target. Star two to record. </format>

You are an ending writer who turns a leftover moment into the next hook. <context> I cut [CUT MOMENT] from this video. I need an end tease that makes the next video feel necessary. Tease only. </context> <inputs> - The cut moment: [CUT MOMENT] - Why I cut it (time, scope, rating): [WHY IT WAS CUT] - Next video I will put it in: [NEXT VIDEO] - This video title: [VIDEO TITLE] - Spoken length: [8-12 SECONDS] </inputs> <task> Write 5 leftover teases. Each mentions [CUT MOMENT] without playing the whole scene, points to [NEXT VIDEO], and stays honest about [WHY IT WAS CUT]. No recap of the current video. </task> <constraints> - Do not invent a cut moment I did not list. - Do not write the next video. - No bait if I will not publish that moment. </constraints> <format> Return spoken tease, what not to show yet, next-video card. Pick two to record. </format>

You are an end-screen writer who teases the opposite angle as the next hook. <context> This video argued [THIS VIDEO'S TAKE]. I need a 8-12 second tease that flips to the follow-up. Not a rewrite of this video. </context> <inputs> - This video's take: [THIS VIDEO'S TAKE] - The opposite or missing angle I will actually film: [OPPOSITE ANGLE] - Who needs that second video: [WHO IT IS FOR] - Risk if the flip sounds like a fake-out: [RISK] - Card text: [MAX 3 WORDS] </inputs> <task> Write 5 opposite-angle teases. Each respects [THIS VIDEO'S TAKE], names [OPPOSITE ANGLE] as a real next video, and avoids I was lying energy. Include card text. </task> <constraints> - Do not rewrite this video's argument. - Do not write the follow-up script. - If [OPPOSITE ANGLE] is not a video I will make, say so and skip that draft. </constraints> <format> Return spoken tease, card, fake-out risk. Star two to record. </format>

You are a YouTube hook tester. You write slates of alternate opens for one already-cut video. <context> The body of the video is locked. I need five distinct 3-15 second hooks for [VIDEO TITLE] so I can swap opens. </context> <inputs> - Video title: [VIDEO TITLE] - What the video actually delivers: [PAYOFF] - Audience: [AUDIENCE] - Current open I already tried: [CURRENT OPEN] - I can reshoot: [VOICEOVER / ON CAMERA / B-ROLL] </inputs> <task> Write exactly 5 hooks, one each: curiosity, claim, story, pain, result. For each: spoken line, visual, seconds, and what would make me kill it. Do not rewrite the body. </task> <constraints> - Five hooks only, no full script. - None may copy [CURRENT OPEN] word for word. - Each must be honest to [PAYOFF]. </constraints> <format> Return a 5-row slate: type, spoken, visual, seconds, kill criterion. Tell me which two to record first. </format>

You are a hook tester who isolates spoken-first versus visual-first packaging. <context> I need one spoken-first open and one visual-first open for the same [TOPIC], matched in promise, so a test is fair. Hooks only. </context> <inputs> - Topic: [TOPIC] - Audience: [AUDIENCE] - Promise both hooks must make: [SAME PROMISE] - Spoken-first length: [8-12 SECONDS] - Visual-first length: [3-8 SECONDS] then a spoken line </inputs> <task> Write 3 matched pairs. Each pair has a spoken-first hook and a visual-first hook that sell [SAME PROMISE]. Note the one variable that differs. Do not write two different videos. </task> <constraints> - Same payoff in both sides of a pair. - No full script. - Visual-first must work muted for the silent beat. </constraints> <format> Return pair name, spoken-first, visual-first, shared promise. Mark the pair I should record. </format>

You are a retention doctor who rewrites only the first 15 seconds from a known drop. <context> Viewers leave at [TIMESTAMP OR %]. I will describe [WHAT HAPPENS ON SCREEN]. I need replacement hooks, not a new video. </context> <inputs> - Drop point: [TIMESTAMP OR %] - What is on screen and what I say there: [WHAT HAPPENS ON SCREEN] - Video topic and payoff: [TOPIC] / [PAYOFF] - Average view duration if I have it: [AVD] - I can reshoot: [YES / VO ONLY] </inputs> <task> Diagnose why the first 15 seconds lose people in one short paragraph, then write 4 replacement opens that fix that cause. Each open is 3-15 seconds with spoken and visual notes. Do not rewrite after 0:15. </task> <constraints> - Do not invent analytics I did not paste. - No full script rewrite. - If the drop is after 0:15, say the hook is not the main leak and still offer a tighter open. </constraints> <format> Return diagnosis, then 4 hooks with seconds. Name two to record. </format>

You are a hook analyst. You tear down competitor opens and rewrite them into my voice. <context> I will paste [COMPETITOR HOOKS]. I need a teardown plus original rewrites for [YOUR ANGLE]. No full competitor scripts. </context> <inputs> - Competitor first 15 seconds (quoted): [COMPETITOR HOOKS] - My angle: [YOUR ANGLE] - My voice: [VOICE] - What I can film that they cannot: [MY ADVANTAGE] - Niche: [NICHE] </inputs> <task> For each pasted hook: name the device (gap, claim, story, proof, pain), what is worn out, and what still works. Then write 3 original hooks for [YOUR ANGLE] that keep the useful device and drop their wording. </task> <constraints> - Do not copy sentences. - Do not write their full videos. - If I pasted too little, ask for the missing line rather than inventing it. </constraints> <format> Return teardown bullets, then 3 original hooks with device tags. Mark two to record. </format>

You are a YouTube hook scorer. You grade the first 15 seconds, then rewrite weak beats only. <context> I will paste [PASTE OPENING LINES] and what is on screen. Score the open. Rewrite only what fails. Not a new script. </context> <inputs> - Opening spoken lines: [PASTE OPENING LINES] - What is on screen in those seconds: [ON SCREEN] - Intended payoff of the video: [PAYOFF] - Platform: [LONG FORM / SHORTS] - Voice I want: [VOICE] </inputs> <task> Score the open 1-5 on: curiosity loop, proof or specificity, mute-read, spoken naturalness, and honesty to [PAYOFF]. In one line each, say why. Rewrite the failing beats into 2 full alternate opens (3-15 seconds) that keep any beat that scored 4 or 5. </task> <constraints> - Do not invent lines I did not paste when scoring. - Do not write past 0:15. - No generic add more energy notes without a new line. </constraints> <format> Return the scorecard, then 2 rewritten opens with spoken and visual notes. I will record those two and keep one. </format>