How to write an Instagram Reels script that holds viewers to the end

The Reel opens well. The first line lands. Then, around second eight, the viewer is already on someone else’s video — and you never got to the tip. Improvising on camera or writing only a hook leaves a hollow middle: filler, extra backstory, a CTA slapped on the last frame.

That cost is not “lack of charisma.” Instagram reads retention. People who leave in the middle third teach the system not to distribute the next video. Without an Instagram Reels script with timestamps, you publish an orphaned hook.

This guide is the spoken piece: hook, body, and CTA in seconds, with copy-paste examples and the mistakes that lose people at the end. It is not how to write the first-three-seconds hook, not Reels cover design, and not hashtags for Reels. It is what you say, beat by beat, until the last second.

What an Instagram Reels script is — and what it is not

An Instagram Reels script is a spoken score. It marks what leaves your mouth (or the voiceover), what appears on screen, and at which second the viewer still needs to be there. It fits in a notes app. You do not need a Hollywood storyboard.

It does not replace:

• The isolated hook. The hook is the gate. The script is the whole hallway. No gate, nobody enters; no hallway, nobody reaches the exit.
• The caption. Captions live after play. Audio viewers already decided whether to stay. Do not mix on-screen text with the feed paragraph.
• The cover. The cover fights for the tap on the grid and in Reels. The script fights for the seconds after the tap.
• The teaching topic. Educational Reels decide what to teach. The script decides in which second each line lands.

Three lines on paper already count as a script:

1. Line 1 (0–3s): tension or a scene.
2. Middle (until ~3s before the end): one rule, one step, one “that’s why.”
3. Close: a single verb — comment, save, go to the bio.

If line 2 is missing, you have a hook with a ask. That wears people out. If line 3 is missing, the video dies on “nice” and nobody does anything.

Instagram’s Help Center on creating a Reel describes the format (vertical, in-app edit, on-screen text). The script is what you bring to that screen, not the record button.

Hook, body, CTA: three blocks on a clock

Think three rooms, not “talk until you run out of air.” The clock changes with length. The logic does not.

Hook. A scene or a claim the viewer recognizes. “This Monday the oven ran late again” beats “today I want to talk about time management.” Meta’s Reels guidance for businesses asks you to earn attention early and keep the video easy to follow — no miracle percentages, just the obvious: if the open is slow, the rest never plays.

Body. One promise kept. If the hook was “the cake splits in the oven,” the body is why and what to do. Do not open a second topic. Do not tell the bakery origin story from 1998. Three short sentences beat a paragraph.

CTA. One verb, one destination. “Comment OVEN,” “save this for the next bake,” “the order menu is in the bio link.” Two CTAs in the same second fight. The viewer does not choose: they leave.

Mark on paper the second each block starts. If you cannot point to second 12 in a 15s Reel, the body ate the close — or the hook ate the body.

Three copy-paste scripts (shoot today)

Swap the name, the product, and the city. Do not swap the skeleton.

1. Small business — bakery in Austin (15s)

On screen: hands slicing a loaf that split down the middle. Text: “stopped splitting.”

• 0–3s: “If the loaf splits in the oven, it is not ‘Austin humidity.’”
• 3–12s: “Crack the door halfway through and the temperature shock splits the crumb. Keep it shut. Open on the timer. Line the pan; don’t flour the sides like a snowstorm.”
• 12–15s: “Save this for the next bake. Weekend order sheet is in the bio link.”

Use: a shop owner who does not want to be on camera the whole time. Hands plus voice is enough.

2. Solo creator — bookkeeper in Denver (30s)

On screen: a spreadsheet cell in red, then her face.

• 0–3s: “Your client dropped the CSV in Slack and thinks the books are ‘already done.’”
• 3–12s: “A file in a folder is not a review. Three checks fit in a Reel: date, category, duplicate. One miss becomes a messy close, not a ‘small detail.’”
• 12–25s: “On Tuesdays I open the file on the call and read the three lines out loud. I don’t send a silent PDF.”
• 25–30s: “Comment CSV if you want the checklist I read on that call. It is not a course — it is the paper I actually use.”

Use: authority without becoming an influencer. The scene is the call, not a generic “get organized with money” tip.

3. Agency — three clients, one method (45s)

On screen: whiteboard with three columns: hook / body / CTA. No client logos.

• 0–4s: “The client asks for ‘a viral Reel.’ We send back a 30-second script.”
• 4–18s: “Column 1: the sentence the owner already typed in iMessage. Column 2: one proof — back-of-house, before/after, a number the shop already has. Column 3: one verb.”
• 18–38s: “We do not film three clients on the same day with the same opener. We swap the product and the neighborhood. The skeleton stays. That is what scales without looking like the same Reel with a new filter.”
• 38–45s: “If you handle a shop and a clinic, DM the word SCRIPT. We send the three-column grid — you fill it in. We do not hit publish for you.”

Use: an agency that needs a process, not “inspiration when the camera is already rolling.”

Shoot all three with the phone on a stand, a timer in view, on-screen text only for keywords (split, CSV, three columns). The rest is speech.

15, 30, or 60 seconds: pick the clock

Length is not vanity. It is how much one idea can carry.

15s when the rule fits in one sentence and one action. “Keep the oven shut.” “Don’t dump the CSV in Slack.” If you need “first… second… third…,” you already left 15.

30s when there is proof. The bookkeeper shows the cell. The bakery shows the loaf before and after. The agency shows the board. Proof without a clock turns into a lecture.

45–60s when the body has three beats: problem, mechanism, example. Past a minute, ask whether this is a carousel. A carousel holds a list; a Reel holds a scene.

Pacing rules, any clock:

• One idea per video. The second idea is the next Reel in the series.
• Cut the silence. “So…” and “like…” are dead seconds.
• On-screen text confirms the line; it does not compete with it. If the overlay says something else, people read and leave.
• Loop: the last frame can return to the first. The CTA still has to exist before the loop — or the ask never lands.

Do not invent “the algorithm prefers 7.3 seconds.” Prefer the length where you can point to hook, body, and CTA on paper.

Common mistakes that lose people in the last third

The last third is where scripts usually fail — not the hook.

1. Hook with no body. “Nobody taught you this” plus 12 seconds of stalling. The viewer waits for the reveal. If the reveal is “post consistently,” they already left.

2. Body that changes subject. You started on the loaf and ended on “follow for more entrepreneurship tips.” Two topics = two videos. On the second topic, retention drops and the CTA never shows.

3. CTA hidden on the last frame. A “link in bio” sticker with no spoken line. Silent viewers might read it. Audio viewers wait for the sentence. Say the verb. Write the verb. One only.

4. An eight-second apology. “Hey guys, I’m Ana, this is my bakery in Austin, before we start I just want to thank…” Instagram already tested whether they stay. You spent the hook on an intro.

5. Five steps in a 15s Reel. Steps 4 and 5 fall off the frame. Three steps want 30s. Five steps want a carousel.

6. Double CTA. “Comment, save, share, and tap the link.” Nobody does four things. Pick what you will count this week: comments (a question), saves (they’ll need it later), clicks (an offer that is ready).

7. Stage-teleprompter sentences. Thirty-word lines. Nobody talks like that on a sidewalk. Cut at the period. Stand up to record. If the sentence does not fit in one breath, it does not fit in the Reel.

None of these get fixed with a new hashtag. Hashtags distribute. Scripts hold. Different jobs.

Batch scripts for the week without sounding like a bot

Daily improv burns you out. Five identical scripts with the noun swapped also sound fake. The middle path is a grid, not a loose generator.

On Monday, open last week’s conversations: the shop’s texts, the consulting call, the client brief. Each complaint becomes a hook. Each fix you already gave becomes the body. The CTA follows what you can actually deliver — a checklist, a menu, a DM.

Three roles, three rhythms:

• Small shop: two 15s Reels (back-of-house + one rule) and one 30s (before/after). Shoot Wednesday after close, same corner of the kitchen.
• Solo creator: one 30s per call theme. Do not film “content.” Film the sentence you repeated to three clients.
• Agency: the three-column grid filled in by the client in writing. You only tune the second and the overlay. The owner films. You do not have to appear.

Contents Pilot drafts the pieces that travel with the video — summary carousel, cover, caption, brand kit — from the same brief. You paste the three script lines into the brief, generate the stills, check the tone, and schedule what belongs on the feed. The product does not publish the Reel for you and does not replace your voice on the audio.

If the generated copy comes out too polished, go back to the scene. “The loaf splits when you crack the oven door” beats “optimize your thermal process.” The script needs the shop slang, the client’s CSV, the agency whiteboard. AI speeds up the visual pack. The spoken line stays yours.

What to check after you publish — without becoming a chart hobbyist

Do not ask a like ranking to judge a script. Ask whether people stayed.

On a professional account, open the professional dashboard and account insights. On the Reel, look at plays, average watch time, and interactions (comment, save, share). Instagram does not always give a second-by-second drop-off map like a long-form editor — so use what is there:

• Lots of plays, low average watch time. The hook earned the tap. The body did not hold. Rewrite the middle, not the cover.
• Average watch time near full length, weak CTA. They watched. The ask was missing or late. Speak the verb three seconds before the end.
• Comments off-topic. The hook promised one thing and the body delivered another. Align the three lines.
• High saves. The body taught something worth keeping. Good signal for the next 30s in the same series — not for a niche change.

A weekly review fits in one line per video: length, average watch time, whether the CTA was spoken, which column of the grid (hook/body/CTA) you would change. Pair that with the metrics that actually matter on the rest of the profile so you do not optimize a Reel as if it were a static post.

When the same middle works twice, do not throw it away. Turn it into Thursday’s carousel and Friday’s Stories with the same rule. The script is the brief. The other formats are cuts.

Want those three paper lines turned into a cover, carousel, and caption in your brand — and a tone check before you hit record? Open Contents Pilot and generate the pack for the next Reel today. You speak on the audio. The product does not publish on its own.