🪄 AI Summary
What You'll Get From This Blog
- 10 ready-to-copy Claude prompts specifically built for YouTube video scripts - no setup needed, just fill and use
- The SCENE Framework - a 5-input system that makes every AI prompt sharper and more conversion-ready
- The real reason your scripts fail - it's not the AI, it's what you're putting inside the brackets
- Prompt-specific failure warnings - each prompt comes with exactly where and why it breaks in real use
- Formats covered - hooks, case studies, thought leadership, framework explainers, podcast clips, webinar repurposing, stat-led scripts, and full end-to-end scripts
- Insider fill-in tips - how to source pain points from sales calls, find proprietary stats, and surface a founder's real contrarian view
- Proof it works - real client results (750K views, 200K+ views) showing what consistent execution with this system actually produces
- Honest bandwidth reality check - an acknowledgment that the prompts are the easy part, and what it actually takes to execute at scale
Most B2B YouTube video scripts die in the first 8 seconds. Not because the topic is wrong, because the opening sentence reads like a press release. You've got a founder with 15 years of hard-won insight, a product that genuinely solves a painful problem, and a video that opens with "Hi, welcome back to our channel." Nobody's staying for that. The brutal truth: 70% of viewers drop off in the first 30 seconds of a YouTube video. The script is almost always why. This blog gives you 10 Claude prompts engineered to fix exactly that, and shows you what most marketers get wrong even after they start using AI.
Read till the end to get most out of this blog.
The Real Problem (And It's Not What You Think)
You're not bad at YouTube because you lack ideas. You're bad at YouTube because you're writing scripts like blog posts with a camera pointed at you.
B2B marketing teams know the script matters. But when a content calendar is stacked, a quarterly campaign is live, and the founder just sent a 47-minute webinar recording asking for "a few short clips," the YouTube script becomes an afterthought. The result: a video that explains features nobody asked about, to an audience that didn't finish the thumbnail, distributed on a schedule that makes no algorithmic sense.
Here's the internal conversation most marketing teams are having: "We post consistently. We have good production. Why is nobody watching?"
The answer is almost always the script. Specifically, the hook, the structure, and the call-to-action. Three things AI can help you improve dramatically, if you know exactly how to prompt it, and exactly where the prompts break down.
Why Generic AI Advice Will Still Leave You with Mediocre Scripts
You've probably already tried asking ChatGPT or Claude to "write me a YouTube script about [topic]." And you got something technically correct, vaguely professional, and completely forgettable.
That's not the AI's fault. That's the prompt's fault.
The advice floating around LinkedIn and YouTube ("use AI to write your scripts!") skips the one thing that matters: context engineering. Generic prompts produce generic outputs. What AI actually needs to write a script that sounds like your brand, hooks your specific audience, and builds toward a conversion, is specificity you've deliberately designed into the prompt.
There's another mistake hiding in plain sight: most AI-generated scripts are written for search, not for a viewer who's already skeptical. B2B buyers watching YouTube aren't passively consuming. They're evaluating. A script that reads like a listicle doesn't move them down the funnel.
At Komet, we've repurposed hundreds of hours of B2B content, podcasts, webinars, keynotes, into short-form video across LinkedIn, YouTube Shorts, TikTok, and Reels. What we've learned is that the clips that drive real pipeline don't come from asking AI to "summarize the key points." They come from a scripting and repurposing framework built around the buyer's moment of decision.
We've also learned exactly where these prompts fail in practice. That's what this blog actually gives you.
How We Build Scripts That Convert
We call it the SCENE Framework, the five inputs every Claude prompt needs to produce a YouTube script worth recording:
- S: Speaker Voice: Who is saying this? What's their authority and personality?
- C: Core Tension: What uncomfortable truth or problem does this video expose?
- E: Evidence Layer: What specific stat, story, or proof point anchors the claim?
- N: Next Micro-Action: What do you want the viewer to do or believe at the end?
- E: Enemy: What common belief or approach does this video contradict?
Every prompt below is built around at least four of these five inputs. Use them in sequence for a single video, or pick the ones relevant to your content format.
One warning before you start: the brackets are traps. [specific pain] filled with "lead generation is hard" produces garbage. [specific pain] filled with "we have three recorded webinars that averaged 8 minutes of watch time each and zero leads attributed" produces a script that actually converts. The difference between a prompt that works and one that doesn't is almost never the prompt structure. It's the quality of what you put inside the brackets. So program your AI wisely and smartly with accurate information of you or your business.
Here’s The 10 Claude Prompts (Easy to Copy, Paste, Record)
Prompt 1: The "Uncomfortable Truth" Hook
Use this for any video where you want to challenge conventional thinking.
You are writing a YouTube script for a [job title] at a B2B [industry] company.
The video is 60-90 seconds. Write a hook (first 15 seconds) that opens with
an uncomfortable truth about [topic] that most people in [industry] quietly
believe but rarely say out loud. Do not start with a question. Do not use
"In today's video." Open mid-claim. Tone: direct, confident, slightly provocative.
End the hook with one sentence that makes it impossible not to keep watching.
Where this breaks: The hook gets watered down in review. A founder or CMO reads the draft, it feels "too bold," and they soften it until it sounds like every other video in the feed. We've seen this happen on at least half the first drafts we produce. If your hook could be said by a direct competitor without changing a single word, it needs to go back. The discomfort is the point.
Note: We run a quick test on every hook before approving it. We ask: would this sentence make someone stop mid-scroll and feel slightly called out? If the honest answer is no, we rewrite it. A hook that makes everyone comfortable converts no one.
Prompt 2: The "Before/After" Structure
For testimonials, case studies, or product demos.
Write a 90-second YouTube Shorts script structured as a before/after story.
Before: [Describe the painful situation, be specific]. After: [Describe the
transformation, use real numbers if available]. The script should feel like
a founder explaining this to a skeptical peer at a conference, not a marketer
pitching a product. End with a single CTA sentence that creates urgency without
sounding desperate.
Where this breaks: The "before" isn't specific enough. "We weren't getting views" is not a before state. It's a vague complaint. The prompt works when the before reads like this: "We were posting two long-form videos per week, averaging 340 views each, and had never seen a single inbound lead attributed to YouTube." That specificity is what makes the after land.
Real fill-in example: When we built content for one of our client, the before was exactly that: a library of solid long-form content, a few hundred views per video, no short-form presence, and an investor conversation that had gone quiet. The after was a YouTube Short that crossed 750,000 views, 600+ new subscribers, and their Head of Global Marketing, describing the result to us as something that "impressed our investors and board." That's a before/after worth scripting. Vague inputs would have produced a forgettable video about the same transformation.
Prompt 3: The "Contrarian Take" Script
Built for thought leadership and personal brand content.
You are ghostwriting a YouTube script for [Name], a [title] in the [industry]
space. Their POV is that [contrarian belief about a widely-accepted practice].
Write a 75-second script that argues this position using one specific story
or data point, acknowledges the strongest counterargument in one sentence,
then lands on a clear conviction. Do not hedge. Do not use "it depends."
The viewer should either strongly agree or want to debate this.
Where this breaks: The founder doesn't actually hold a contrarian view. They think they do, but when you ask them to articulate it, what comes out is a mildly-differentiated take that's been smoothed into acceptability by years of investor pitches and LinkedIn posts. Claude will write a technically convincing script around a weak opinion, and it will fall completely flat on camera because the speaker doesn't believe it enough to deliver it with conviction.
Note: We never take a founder's brief at face value for this prompt. We spend 15 minutes in a conversation asking: "What do most people in your space believe that you think is wrong, and what did you learn the hard way that changed your mind?" The real contrarian view almost always surfaces in that conversation, not in a written brief. We use that raw material as the prompt input.
Prompt 4: The "Problem-First" Framework Explainer
For educational content that builds authority.
Write a YouTube script (2 minutes) that explains [framework or concept] to
a [specific role]. Start with a problem they've probably had this week,
make it visceral and specific. Then introduce the framework as the solution.
Use plain language. No jargon. Break the framework into exactly 3 steps.
Each step should have one concrete example. End with what changes for them
if they implement this today.
Where this breaks: The framework has too many steps, or the steps aren't genuinely distinct from each other. We worked with a B2B SaaS client whose internal framework had seven stages. Every stage was real and considered. But at step four, viewers tuned out, regardless of how good the content was. We stripped the framework to three before the prompt produced anything usable. If you can't explain your framework in three steps that a smart person could repeat from memory, the problem isn't the prompt — it's the framework.
Note: We enforce a "three-step ceiling" on every framework explainer we produce. If a client insists on more, we build a series: three steps per video, published across three weeks. That approach generates more views, more return viewers, and more pipeline than any single seven-step explainer ever has.
Prompt 5: The "Podcast Clip Repurpose" Prompt
For turning existing audio into scripted video.
Here is a transcript excerpt from a podcast episode: [paste 200-400 words].
Identify the single most tension-filled or insight-dense moment in this
transcript. Rewrite it as a standalone 60-second YouTube Short script.
Add a hook sentence at the start that gives a viewer with no context a
reason to watch. Do not reference "the podcast." The clip should feel
like an original, standalone piece of content.
Where this breaks: Two things kill this prompt. First, pasting too much transcript — anything over 400 words and Claude loses the thread, produces a summary instead of a sharp script, and the output reads like a highlight reel rather than a standalone piece. Second, pasting a raw, uncleaned transcript. Filler words, crosstalk, repeated false starts, and tangents confuse the model's sense of what the "insight-dense moment" actually is. We never paste raw transcript into this prompt.
Note: Before we touch this prompt, we clean the transcript: filler words removed, tangents cut, crosstalk stripped. Then we paste no more than 300 words representing the section we've already identified as the strongest moment. We find the moment first. Then we let Claude script it. Letting Claude find the moment from a raw 45-minute transcript is asking it to do editorial work it isn't built for.
💡 Pro tip: This prompt pairs directly with Komet's Complete Podcast Repurposing Guide, a free walkthrough of exactly how we turn every episode into a multi-platform content asset that generates pipeline long after it publishes.
Prompt 6: The "Stat-Led" Authority Script
For data-heavy content or research findings.
Write a 90-second YouTube script that opens with this statistic: [stat].
Explain in plain English why this number should concern a [specific role].
Then provide one tactical action they can take this week based on it.
Tone: senior strategist briefing a peer, not a thought leader on a TED stage.
End with a question that makes them want to comment.
Where this breaks: The stat is too well-known. "70% of B2B buyers don't talk to sales until they've already made a decision" has been in every B2B marketing deck since 2015. When a viewer has already heard the stat a dozen times, the script loses its authority signal before the second sentence. The prompt performs best with proprietary data, category-specific research from a niche industry report, or a stat your team discovered internally that nobody else is citing yet.
Note: We ask every client before using this prompt: "Do you have any data from your own customers or platform that you've never published?" That's almost always where the best stat is hiding. A number that only your company can cite is worth ten times a recycled industry statistic.
Prompt 7: The "Enemy Framing" Script
For positioning against a category or approach, not a competitor.
Write a YouTube Short (60 seconds) that positions [your approach] as
the alternative to [common but ineffective approach]. Don't name any
company. Frame the "enemy" as an outdated method or mindset. Open with
a specific scenario where that outdated approach fails. Pivot to the
better way. End with a clear POV statement the viewer can repeat to
their team.
Where this breaks: The enemy is too abstract. "Old-school marketing" is not an enemy. "Traditional demand gen thinking" is not an enemy. An enemy has to be specific enough that a viewer watching your video thinks "that's exactly what we've been doing." The moment of recognition is what makes this prompt work. Without it, you've written a vague criticism that offends no one and converts no one.
Note: We find the enemy by asking clients one question: "What's the last piece of content your team produced that completely flopped, and what was the assumption behind it?" The answer to that question is almost always the enemy. In one case, a client described uploading a full 45-minute webinar to YouTube and calling it "content distribution." That became the enemy in a video that performed 8x better than anything else they had posted that quarter.
Prompt 8: The "Webinar Clip Extraction" Script
For marketing teams sitting on recorded webinar content.
Here is a webinar transcript section: [paste]. This is a 45-minute webinar
hosted by [name] on [topic]. I need a 75-second YouTube Short script based
on the most shareable insight in this excerpt. The script should work without
any slides or screen shares, voice and face only. Rewrite for spoken delivery.
Cut anything that only makes sense in the context of the full webinar.
Add a one-sentence closing that directs viewers to [action/resource].
Where this breaks: The webinar was slide-heavy, and the audio alone doesn't hold up. A speaker saying "as you can see here" or "this chart shows" is completely unusable for short-form. We've opened webinar transcripts where 60% of the speaker's value was in their slides, not their words. No prompt can fix that. The only clips worth extracting are moments where the speaker drops into a story, shares a direct opinion, or says something that would make a peer lean forward.
Nnote: Before we run this prompt, we scan the transcript for one thing: moments where the speaker uses the word "I" in a personal story or "you" in a direct challenge to the audience. Those are the extractable moments. Everything else is context that only made sense in the room. One of the SaaS companies came to us with exactly this challenge across their webinar and podcast — we built a system for identifying those moments at scale, and it's how they now consistently ship content without having to hand-hold the editing process.
Prompt 9: The "ICP Pain Point" Direct Script
For bottom-of-funnel content targeting a specific buyer.
Write a 90-second YouTube script targeting a [role] at a [company size]
[industry] company. Their biggest frustration right now is [specific pain].
The video should demonstrate that we deeply understand this frustration
in the first 20 seconds, offer one clear perspective shift in the middle,
and end with a specific next step (not "learn more"). Do not pitch a product.
Build trust first.
Where this breaks: The pain point is too broad. This is the most common failure mode we see across every client vertical. "Lead generation is hard" is not a pain point — it's a category. A pain point sounds like this: "We have four webinar recordings from last quarter sitting in a Google Drive. Each one got maybe 80 live attendees. We've done nothing with them since. Our content calendar is empty and the next webinar isn't for six weeks." That is a pain point. The more your [specific pain] reads like something a real buyer said in a sales call, the better this prompt performs.
Note: We pull our pain point inputs directly from client sales call recordings or onboarding calls whenever possible. The exact language a buyer used to describe their problem is almost always sharper than anything a marketing team would write in a brief.
Prompt 10: The "Full Script with SCENE Framework" Prompt
The master prompt: use this when you need a complete, polished script.
Write a complete YouTube script (90-120 seconds) using this framework:
Speaker Voice: [Name, role, what makes them credible on this topic]
Core Tension: [The uncomfortable truth this video is built around]
Evidence Layer: [One stat, case study, or story that proves the tension is real]
Next Micro-Action: [What the viewer should do or believe after watching]
Enemy: [The common approach or belief this video challenges]
Format: Hook (15 sec) / Problem (20 sec) / Insight-Turn (40 sec) /
Proof (20 sec) / CTA (10-15 sec).
Tone: [Direct / Conversational / Authoritative].
Do not use filler phrases. Every sentence must move the viewer forward.
Where this breaks: All five inputs are filled with safe, corporate language. When the Enemy field reads "lack of content strategy" and the Core Tension reads "content marketing is evolving," the output will be technically structured and completely unmemorable. The SCENE Framework only produces a great script when the inputs are sharp enough to make someone slightly uncomfortable.
Note: We rewrite the Enemy field at least three times before approving it for this prompt. The first version is always too polite. The second is getting warmer. By the third, it usually names something specific and real that a prospect would recognize immediately. That's when we use it.
What Happens When You Actually Use The System
Goliath.io came to Komet sitting on a library of long-form content and getting a few hundred views per video. No real distribution strategy. No repurposing system. Their content was solid. The problem was delivery and format.
After building a structured short-form video workflow, pulling clips from existing recordings using a prompt-and-repurposing approach similar to what you've just read, one of their YouTube Shorts crossed 750,000 views. They added over 600 new subscribers directly from short-form. Ariana Faustini, Head of Global Marketing, called it out specifically: "This has impressed our investors and board."
That result didn't come from one great script. It came from a repeatable system, applied consistently, with practitioner judgment on what to extract and how to frame it.
Similarly, Walden Catalyst Ventures reached 200,000+ views on YouTube Shorts by turning their existing content into a consistent short-form publishing engine, without recording anything new.
Both of these results share one root cause: they stopped treating scripting as a one-off task and started treating it as a production system. The prompts are the starting point. The system is what compounds.
You've Got the Prompts. Here's the Harder Question.
These 10 prompts will make your next YouTube script better.
But here's the honest follow-up: the bottleneck for most B2B teams isn't knowing how to write scripts, it's having the bandwidth to actually execute consistently. One good script takes 45-90 minutes to write, refine, and approve, not counting the editorial judgment required to find the right moment in a 47-minute recording in the first place. Multiply that across 8-12 videos a month and you've consumed a significant chunk of your content team's capacity before a single frame is recorded.
That's exactly what Komet solves. We take your existing podcasts, webinars, and long-form recordings and build a full short-form content engine, from clip identification and scripting to platform-ready delivery across LinkedIn, YouTube Shorts, TikTok, and Reels.
You've already created the content. We make it work harder for you or/and your business.
Start with our Complete Podcast Repurposing Guide, a free, no-fluff walkthrough of how to turn every episode into a multi-platform content asset that keeps generating pipeline long after it publishes.
Or if you're ready to stop doing this manually:
See Exactly How We Would Build Video Content Engine
Book Your Free Strategy Call. In 15 minutes, we'll show you specifically how your existing content could be turning into qualified pipeline, with real examples from brands in your space.
Spots are limited to 5 calls per week. We only work with brands who is serious about the growth and genuinely wants to make an impact and yours probably is.
Author:
Rajan Soni
Rajan is passionate about marketing & business. He believes in process & preparation over everything else.
