Drama Ads: The Claude Scripting Guide

This guide gets Claude to write drama ads: 2–7 minute vertical episodes where the story earns the attention and the product becomes the turning point.

You get the 6-beat structure, the rules for placing the product inside the plot, conflict templates for four categories, the cliffhanger formula, and four prompts that go from a one-line idea to a shot-by-shot script.

Use Claude Fable 5.1 or Opus 5.5. Both write strong dialogue, and Fable is noticeably better at the first three seconds — which in this format is everything.

The last part covers getting the script on screen. A script is only half an ad, and that's where most teams stall.

People open Instagram and TikTok to watch something, not to be sold to. A drama ad gives them the thing they came for, then uses the goodwill to introduce the product.

Three mechanics do the work:

• Open loops hold attention. A question the viewer needs answered — will she quit, who sent the message — keeps them watching far past the point a normal ad loses them.
• The product arrives as relief, not interruption. When it solves a problem the viewer already cares about, it reads as plot, not pitch.
• Episodes compound. A viewer who watched episode 1 seeks out episode 2. That's an audience you built, not one you rented for 15 seconds.

The format borrows from vertical micro-drama apps, which trained a large audience to watch short, high-stakes episodes on their phones.

When not to use it

• Your product can't plausibly change a situation. Drama needs stakes the product resolves. A commodity with no emotional job struggles here.
• You can't produce more than one episode. One-off dramas work, but the format pays most as a series.
• You need volume tomorrow. A drama episode takes longer to write and make than a UGC clip. Run it alongside fast formats, not instead of them.

Every drama ad that holds attention follows the same six beats. Timings below are for a 3-minute episode; stretch or squeeze them in proportion for 2 or 7.

• 1 — Beat: Cold open · Time (3-min): 0:00–0:05 · Its job: Drop into the most charged moment, mid-scene, no setup · The viewer feels: "Wait — what's happening?"
• 2 — Beat: The wound · Time (3-min): 0:05–0:30 · Its job: Who she is, what she wants, who's standing in the way · The viewer feels: "I know someone like her"
• 3 — Beat: The squeeze · Time (3-min): 0:30–1:30 · Its job: Three blows, each worse than the last · The viewer feels: "This isn't fair"
• 4 — Beat: The turn · Time (3-min): 1:30–2:00 · Its job: She makes a choice — and the product is part of it · The viewer feels: "Okay, here we go"
• 5 — Beat: The payoff · Time (3-min): 2:00–2:40 · Its job: The reversal: she wins, the antagonist doesn't · The viewer feels: Satisfaction
• 6 — Beat: The hook out · Time (3-min): 2:40–3:00 · Its job: A new question, then a soft call to action · The viewer feels: "I need episode 2"

Three rules hold across all six.

The cold open is a flash-forward. Take the most emotional frame from beat 3 or 5 and put it first. Then cut to "Three weeks earlier." The viewer stays to find out how she got there.

Beat 3 is where the money is. Most weak drama ads rush the squeeze. Three escalating blows — a rival, then a public humiliation, then a personal cost — is what makes the payoff land.

Nothing in beats 1–3 mentions the product. The viewer has to care about her before they'll care about what she uses.

The product should change what happens next, not pause the story to explain itself. Four rules get it there.

1. Plant it early, use it late. Show the product once in beat 2, in the background, unexplained — on her desk, in her bag. When it matters in beat 4, the viewer recognises it. That feels like story, not insertion.
2. She chooses it. The product is never handed to her as a fix. She decides to use it, as part of deciding to fight back. The choice is the character moment; the product rides along.
3. Show the effect, not the features. Nobody in a drama says "clinically proven niacinamide." They say "you look different" — and she smiles. The benefit is what other characters notice.
4. One line, maximum. If the product needs explaining, one character asks and another answers in a single line. Anything more and the episode turns into an ad in the middle.

Three ways to land it

• The tool — How it works: She uses it to win the next scene · Fits: Apps, tools, productivity, finance
• The transformation — How it works: It changes how she looks or feels, and others notice · Fits: Skincare, fashion, fitness, supplements
• The secret weapon — How it works: Rivals wonder how she did it; the reveal is the product · Fits: Anything with a "how did you do that?" result

Pick one per episode. Mixing them muddies the turn.

Each template is a complete premise: a protagonist, the pressure on her, the wound, how the product turns it, and the payoff. Swap names, settings and the product; keep the shape.

• Skincare — Protagonist: Junior associate at a law firm · Pressure: A senior partner who dismisses her · The wound: Overhears "she looks exhausted, not partner material" · The turn: Rebuilds her morning routine before the big pitch · The payoff: Leads the pitch; the partner asks what changed
• Skincare — Protagonist: Bride three weeks out · Pressure: A future mother-in-law who notices everything · The wound: Engagement photos come back and she hates them · The turn: Commits to a routine, privately · The payoff: Wedding morning; the mother-in-law goes quiet
• Supplements — Protagonist: Single dad, night-shift nurse · Pressure: A custody hearing and constant exhaustion · The wound: Falls asleep at his daughter's recital · The turn: Changes how he fuels his shifts · The payoff: Front row at the next one, wide awake
• Supplements — Protagonist: Former athlete, now a desk manager · Pressure: A younger colleague who outruns him at the company 10K · The wound: Finishes last, in front of his team · The turn: Starts training properly · The payoff: Rematch next year; he doesn't finish last
• Fashion — Protagonist: Designer passed over for a promotion · Pressure: A rival who presents her ideas as his own · The wound: Turns up underdressed to the meeting where he steals her pitch · The turn: Shows up to the next one dressed like she owns the room · The payoff: Pitches her own work; the room listens
• Fashion — Protagonist: Woman newly single after ten years · Pressure: An ex posting a new life online · The wound: Realises nothing she owns feels like her · The turn: Rebuilds her wardrobe, piece by piece · The payoff: Runs into the ex; he barely recognises her
• Apps — Protagonist: Café owner drowning in admin · Pressure: A chain opening across the street · The wound: Misses a supplier payment, loses her best supplier · The turn: Moves her operations onto the app · The payoff: Busiest weekend she's ever had; the chain's queue is shorter
• Apps — Protagonist: Freelancer chasing late invoices · Pressure: A client who keeps "forgetting" to pay · The wound: Can't make rent despite a full calendar · The turn: Automates invoicing and reminders · The payoff: Paid on time, fires the client

Rules for characters

• Give the protagonist a flaw. Pride, avoidance, a tendency to give in. Viewers trust a character who is partly responsible for her own mess.
• Give the antagonist a reason. The best rivals are right about something. A cartoon villain makes the win feel cheap.
• One protagonist per series. Recurring characters are what make episode 2 findable.
• Write specifics, not types. "A senior partner who calls her 'kiddo' in meetings" beats "a mean boss."

Close the episode's question, then open a bigger one in the final ten seconds. The formula is four parts, in order:

1. The win lands. Let the payoff breathe for a beat. The viewer needs the satisfaction first.
2. Something new arrives. A message, a person in the doorway, a line overheard.
3. A face reacts. Cut to the protagonist's expression. Don't show what she sees — show what it does to her.
4. Hard cut. No resolution, no voiceover. The last frame carries the soft CTA: "Part 2 on our page."

Four cliffhangers that work

• The reveal — Example last line: "You don't know who your father really is." · Best after: A personal win
• The strike-back — Example last line: The rival forwards her an email: "Thought you should see this." · Best after: A professional win
• The arrival — Example last line: The ex walks into the café. · Best after: A fresh start
• The offer — Example last line: "We want you in New York. Monday." · Best after: Any win; it raises the stakes

Running it as a series

• Plan three to five episodes before you shoot one. Each is a complete story; together they're an arc.
• Every episode stands alone. A cold viewer of episode 3 must understand it in five seconds.
• Serve episode 2 to people who watched episode 1. Build a custom audience from viewers of 50% or more of episode 1 and run episode 2 to them. You're paying to reach people who already asked for it.
• Keep the product turn in every episode, but in a different scene each time. Repetition of the story beat is fine; repetition of the shot is not.

Create a Claude Project for each series and paste the show bible below into its instructions. Every prompt in the next section then inherits it, so characters, tone and product rules stay consistent across episodes.

Rule 8 matters more than it looks. It keeps scripts cheap to film and simple to generate, which is the difference between a series you finish and a pilot you abandon.

Run these in order inside the Project. Each one takes the output of the last. Read every output before moving on; the first two are where you steer.

1. Loglines

Pick one. If none grabs you, say why and ask for 10 more. Specific feedback ("too corporate", "stakes too small") sharpens the next batch fast.

2. Beat sheet

This is the cheapest place to fix the story. Changing a beat here costs a sentence; changing it after the script costs a rewrite.

3. Script

4. Shot list

Bonus: three cold opens to test

The cold open decides whether anyone sees the rest. Testing three versions of the first five seconds is the highest-leverage thing you can do with a finished episode.

This is the skincare template from the law firm, run through all four prompts. The product is an invented serum, Lumen, in an amber dropper bottle. Length: 3 minutes.

Logline: A junior associate overhears she's "not partner material" and has three weeks to prove the senior partner wrong before the biggest pitch of the year.

The script

Notice what the product does and doesn't do. It's on screen in four places. It has one line — "New routine." Nobody names an ingredient. And the turn is her choice: resetting the alarm, not the bottle, is the moment she fights back.

The first shots of the shot list

• 1 — Time: 0:00–0:05 · Shot: Low angle, medium, slow push in on Grant · What happens: Grant sets down his pen · Dialogue / subtitle: "Where did this come from?"
• 2 — Time: 0:05–0:08 · Shot: Black frame · What happens: Title card · Dialogue / subtitle: [Three weeks earlier]
• 3 — Time: 0:08–0:14 · Shot: Desk height, wide; folders fill the lower third · What happens: Maya working late, lobby empty · Dialogue / subtitle: —
• 4 — Time: 0:14–0:17 · Shot: Insert, close; shallow focus · What happens: Amber bottle behind the monitor, unopened · Dialogue / subtitle: —
• 5 — Time: 0:17–0:24 · Shot: Over Maya's shoulder; partners soft in background · What happens: Grant and Elena pass · Dialogue / subtitle: "Harlow pitch. Maya?" / "Not partner material."
• 6 — Time: 0:24–0:29 · Shot: Close-up, static · What happens: Maya's eyes close, jaw tight · Dialogue / subtitle: —

Shot 6 is the frame from the stills in the post. It's the emotional anchor of the whole episode, and it's also the strongest alternative cold open.

A finished script is where most teams stall. You have two ways to get it on screen.

• What you need — Film it: Actors, a location per scene, a cinematic lighting setup · Generate it: The shot list, a product URL, one character reference
• Time per episode — Film it: A shoot day plus editing · Generate it: An afternoon
• Best for — Film it: Brands with a production budget and a recurring cast · Generate it: Testing premises before committing to a shoot
• Watch out for — Film it: Cost of reshoots when a beat doesn't work · Generate it: Keeping the protagonist consistent across shots

Most teams should generate first, find the premise that works, then decide whether it's worth filming.

Generating it

This is what HeyOz is built for. Paste your product URL and it locks your product — the bottle, the label, the colours — so the amber dropper in shot 4 is the same bottle in the payoff. Then you render the episode shot by shot.

Step 1: Turn the shot list into prompts. Back in your Claude Project:

Step 2: Make the character reference. Generate shot 6 — the close-up of Maya — as a still first. Approve it. That still is the reference every other shot of her is built from.

Step 3: Render on HeyOz. Add your product URL and upload the character still. Paste the prompts shot by shot. Render the product shots first; they're the ones that have to be exactly right.

Step 4: Compare models on the hard shots. Emotional close-ups and the product insert are where models differ most. Render those two on more than one video model side by side and keep the best.

Step 5: Assemble. Cut the shots in order, add the subtitles from the script, and lay a quiet score under it. Silence in the cold open often hits harder than music.

The cold-open variants from the bonus prompt cost one extra shot each. Render all three while you're there.

Judge a drama ad on whether people stay, not whether they click in the first ten seconds. Clicks come late in this format, after the payoff.

Cut three versions from every episode

• Full episode — Length: 2–7 min · What it's for: The main ad; builds the audience for episode 2
• Story cut — Length: 45–60 sec · What it's for: Cold open, wound, turn, payoff; drops most of the squeeze
• Teaser — Length: 10–15 sec · What it's for: Cold open plus the cliffhanger line; sends people to the full episode

The metrics that matter

• Hook rate — How to read it: 3-second views ÷ impressions · Tells you about: The cold open
• Hold rate — How to read it: ThruPlays ÷ 3-second views · Tells you about: Whether the story holds past the open
• Watch-time drop-off — How to read it: Where the retention curve falls in Ads Manager · Tells you about: Which beat is losing people
• "Part 2?" comments — How to read it: Count them · Tells you about: Whether you have a series
• Cost per result — How to read it: As normal · Tells you about: Whether the story sells

How to run the test

1. Put the three cold-open variants in one ad set, same body, same audience. The winner is the one with the best hook rate and hold rate — a cold open that tricks people into staying three seconds isn't a win.
2. Read the retention curve on the winner. A sharp drop at one beat means rewrite that beat, not the whole episode.
3. Make episode 2 when the comments ask for it and your 50%-viewer audience from episode 1 is large enough to serve. Until then, iterate episode 1.

Every variant you test is another render. If a beat needs rewriting, change the shot list in Claude and re-render just those shots on HeyOz — the product and character stay locked, so the new shots cut straight back into the episode.

• People leave in the first 3 seconds — Likely cause: The cold open is setup, not a charged moment · Fix: Re-run the bonus prompt; flash-forward the payoff instead
• Retention falls off in the middle — Likely cause: The squeeze is one blow repeated, not three that escalate · Fix: Ask Claude to make each blow cost her something different: status, money, a relationship
• Comments call it an ad — Likely cause: The product arrived before the viewer cared, or was explained · Fix: Move the plant to beat 2, cut its lines to one
• Dialogue sounds like a brand — Likely cause: Characters are explaining their feelings · Fix: Add to the script prompt: "Nobody says how they feel. Show it."
• The protagonist looks different shot to shot — Likely cause: Prompts describe her differently, or no reference still · Fix: Paste her description verbatim into every prompt; build every shot from the approved still
• The product looks different in the payoff — Likely cause: The product wasn't locked from a reference · Fix: Render product shots from your product URL, not a description
• High hold, no sales — Likely cause: The story works; the turn doesn't connect to buying · Fix: Make the effect visible on screen, and end with a clear link to the product page

A repeatable way to make the format brands are scaling with: a show bible, four prompts, and an episode you can test this week.

Claude writes the show. HeyOz puts it on screen with your product in every frame where it belongs. You decide which episode runs next.

Stop interrupting the show. Make the ad the show.