Skip to content

$ cat prompt.txtYouTube

Viral YouTube Script Builder — Hook + Story + CTA in 30 Min

Full YouTube script with a tested 8-second hook, retention beats every 60 sec, and a high-CTR title set.

featured 30–45 min intermediate Claude Opus, Claude Sonnet, GPT-5#youtube#scriptwriting#viral-hooks
intro.md

Three things make a YouTube video pop: a hook that earns the first 8 seconds, retention beats that prevent the 30-second cliff, and a title that out-clicks the rest of the suggested-video sidebar. This prompt produces all three — a full script with explicit retention structure, plus 10 title variants ranked by predicted CTR.

$ cat prompt.txt | pbcopy

The prompt — copy & paste

prompt.txt
You are a top-1% YouTube scriptwriter who has written for channels with 10M+ subscribers across tech, finance, and education niches. Your job is to write a high-retention, high-CTR script for the video below.

VIDEO INPUTS:
- Topic: {{video_topic}}
- Channel niche: {{channel_niche}}
- Audience skill level (beginner / intermediate / advanced): {{audience_skill_level}}
- Target video length: {{target_length}} (e.g. "8-10 minutes")
- My unique POV / why I'm the right person to make this video: {{your_pov}}

EXECUTE — produce the following 4 deliverables in order:

DELIVERABLE 1 — HOOK (first 8–15 seconds)
Write 3 hook variants, each using a different proven formula:
A. CONTRARIAN — "Everyone says X. Here's why X is wrong."
B. STAKES — "If you do X, you'll lose Y. Here's what to do instead."
C. CURIOSITY GAP — "I tried X for 30 days. The result wasn't what I expected."

Each hook must:
- Be 25-45 words (8–15 seconds when read at natural pace)
- Promise a specific payoff the viewer gets if they watch
- NOT be a "Welcome back to my channel" intro
- Use sentence fragments and rhythm — read it aloud, it should sound like speech, not writing

DELIVERABLE 2 — FULL SCRIPT
Write the complete script structured in numbered "retention beats" — every 45–60 seconds there must be a new beat that gives the viewer a reason to keep watching. Format:

[BEAT 1 — 0:00-0:15] HOOK
(use Hook variant A above)

[BEAT 2 — 0:15-0:45] PROMISE EXPANSION
Briefly preview the 3-4 specific things they'll learn. Tease the BEST one for the end of the video.

[BEAT 3 — 0:45-1:45] MAIN POINT 1
(structured teaching block — concept, example, application)

[BEAT 4 — 1:45-2:00] PATTERN INTERRUPT
A change of pace — a "but here's the thing" moment, a quick aside, a relevant data point.

[BEAT 5 — 2:00-3:00] MAIN POINT 2
(structured teaching block)

[BEAT 6 — 3:00-3:15] PATTERN INTERRUPT
...

Continue with 3-5 main points + interrupts depending on {{target_length}}. End with:

[CLOSING BEAT — last 30 sec]
- The BEST point (the one teased in Beat 2)
- A clear next-action CTA (one specific thing the viewer should do — comment with X, watch the next video, subscribe for the series)

Include B-ROLL CUES in brackets: [B-ROLL: hand pointing to screen showing AliExpress search] — gives the editor explicit shot direction.

Match my {{your_pov}} throughout — every claim should feel like it's coming from MY angle, not generic YouTube voice.

DELIVERABLE 3 — TITLE VARIANTS (10 titles, ranked)
Generate 10 title variants for this video. Each must be:
- 50-65 characters (fits without truncation on mobile)
- High-CTR pattern (curiosity gap, contrarian, number, "how I...", "X without Y", etc.)
- Search-volume aware — include the primary keyword in titles where it fits naturally

After the 10, RANK them in this format:
| Rank | Title | CTR Pattern | Predicted CTR | Search Volume |
The top-ranked title should optimise for the right balance of click-through (suggested-video feed) AND search discoverability.

DELIVERABLE 4 — THUMBNAIL CONCEPT
Suggest 3 thumbnail concepts that match the top-ranked title. For each:
- The visual focal point (face expression + body posture + setting)
- The 3-5 word text overlay (HUGE on mobile)
- A color contrast strategy (most YouTube thumbnails are red/yellow — what would make yours stand out?)

Be SPECIFIC. Generic answers ("your face looking shocked") are useless — I need "left half: my face with mouth open in laughter pointing at screen. Right half: bold yellow text 'I WAS WRONG' on dark background."

$ variables_to_fill_in

{{video_topic}}{{channel_niche}}{{audience_skill_level}}{{target_length}}{{your_pov}}

$ man playbook

Step-by-step playbook

How to actually use this prompt for the best results.

  1. 01

    Pre-fill the variables with brutal honesty

    Especially {{your_pov}} — if you don't have a unique POV on this topic, the script will read like every other YouTube video. Either find your POV first or pick a topic where you actually have one.

  2. 02

    Run the prompt and read all 3 hooks aloud

    Pick the one that sounds most natural in your own voice. Reading aloud catches awkward phrasing the model can't.

  3. 03

    Time-check the script

    Read the full script aloud at natural pace. Most scripts come out 15-30% longer than the {{target_length}} target. Cut accordingly — pacing is the #1 retention lever.

  4. 04

    Pick the title + film a matching thumbnail

    The title and thumbnail must match. Don't pick a curiosity-gap title with a generic thumbnail. The ranked CTR table is your guide — but always trust your knowledge of your own audience over the model's prediction.

  5. 05

    Edit with the B-roll cues as a checklist

    The bracketed [B-ROLL] cues are explicit shot direction. Insert each one during editing. Videos with B-roll at every retention beat outperform raw talking-head videos by 30-50% in average watch time.

$ man tips

Pro tips for better output

  • If you want a Shorts script instead of long-form, run the same prompt with {{target_length}} = '45 seconds' — the structure adapts to a 3-beat format automatically.
  • Save your strongest hook variants — over time you'll build a personal hook library that gets better than the AI's defaults.
  • Re-run the title generator a second time after editing — sometimes the best title is suggested by the actual content of the video, not the original brief.
  • For series content, ask the model in a follow-up: 'now write a Part 2 script that picks up where this video left off and assumes the viewer watched Part 1'. Series videos get 2-3x the watch time per viewer.

$ echo $YIELD

What you'll get

output.md

A complete YouTube script (full word-for-word text with retention beats and B-roll cues), 3 hook variants, 10 ranked title options, and 3 thumbnail concepts. Edit-ready.

$ man faq

FAQ

Will YouTube penalise AI-written scripts?

No — YouTube's policy is about misleading content and synthetic-presentation disclosure (fully AI-generated videos with a synthetic voice/face must be flagged). AI-assisted scripts performed by a human presenter are entirely fine and extremely common, including on top-10-subscriber channels.

Can I use this for faceless / narration channels?

Yes — add 'this is a narration-only faceless channel' to {{your_pov}}. The model will adjust the script structure (more visual cues, less first-person, denser pacing) and the thumbnail concepts (no face).

How long until videos using this prompt actually perform?

Script quality affects retention (watch time per viewer) immediately. CTR-driven growth takes 3-6 months of consistent uploads as the algorithm learns your channel. Both compound — a channel uploading 8 well-scripted videos/month typically sees meaningful subscriber growth within 90 days.

$ ls /prompts/youtube

Related prompts