You open the app to post, the idea in your head sounds great, and then nothing happens. The idea never turns into a video because "make a TikTok about X" isn't something you can film. A script is.
A short-video script isn't an essay with a camera pointed at it. It's a short list of beats, and each beat has three layers: what you say, what's written on screen, and what the viewer sees. This post walks through how to get from a rough idea to that list, using a simple four-beat template: HOOK → BUILD → PAYOFF → CTA.
What a filmable script contains
If you can hand your script to a friend and they could film it without asking you questions, it's filmable. That means every beat has:
- Timing: roughly where it starts and ends (0–3s, 3–10s…).
- Voiceover: the exact words, written to be said out loud.
- On-screen text: the words overlaid on the video, readable with the sound off.
- Visual or shot cue: what's in frame: your face, hands, a screen recording, a B-roll clip.
Most scripts people write only have the voiceover. The other two layers get made up during filming, which is why shoots take longer than the video they produce.
Step 1: Turn the idea into one angle
A rough idea is a topic. An angle is the single sharpest thing the video will say about that topic. Write it as one sentence.
Example (made up for this post):
- Rough idea: "something about houseplants and travel"
- Angle: "Your plants don't die while you travel because you forgot them. They die because you watered them too much right before you left."
The angle decides everything after it. If you can't say it in one sentence, you probably have two videos.
Step 2: Write three hooks, then pick one
The first few seconds decide whether anyone sees the rest. Instead of polishing one opener, write three in different styles and choose the strongest:
- Mistake: "You're killing your plants the night before your trip."
- Question: "Why do plants die when you're away, even if you watered them?"
- Result first: "Two weeks away, zero dead plants. Here's the setup."
Pick the one that matches the payoff most closely. Hook 1 fits our angle best, because the video is about one specific mistake. Keep the other two. They're your A/B tests for a repost later.
Step 3: Fill in the four beats
Here's the template for a video of about 30 seconds. Stretch or shrink the middle for other lengths. Keep the hook short.
| Beat | Time | Job |
|---|---|---|
| HOOK | 0–3s | Stop the scroll and promise something specific |
| BUILD | 3–12s | Name the problem or tension, and why it matters |
| PAYOFF | 12–25s | Deliver what the hook promised: the fix, steps, or reveal |
| CTA | 25–30s | One clear ask: follow, save, comment, or visit |
Two rules make this work:
- The payoff has to match the hook. If the hook says "the one mistake", the payoff is that mistake and its fix, not a list of ten tips.
- One CTA only. "Follow, like, comment and check the link" means none of them happen.
Step 4: Write voiceover you can actually say
Writing for the ear is different from writing for the page. A few checks:
- Read every line out loud. If you trip over it, rewrite it.
- One breath per line. Long sentences become two lines.
- Use contractions and plain words. "You're" not "you are", "use" not "utilize".
- Cut the warm-up. "Hey guys, so today I wanted to talk about…" spends your hook on nothing.
- Say the specific noun. "Your monstera" lands better than "your plants" if that's what's in frame.
Step 5: Add on-screen text and a shot for every beat
Now go beat by beat and fill in the other two layers.
On-screen text should make sense without audio. It doesn't need to repeat the voiceover. It often works better as a short label: "The night-before mistake", "Do this instead", "Step 2".
Shot cues should be concrete enough to film: "close-up of soggy soil, finger pressing in", "screen recording of a calendar reminder", "face to camera, holding the pot". If a beat has no visual idea, that's a sign the beat is too abstract.
A full worked example
Here's the houseplant script filled in. It's an example written for this post, not real data, so check any plant-care claim against a source you trust before you film something like it.
| Time | Voiceover | On-screen text | Shot |
|---|---|---|---|
| 0–3s | "You're killing your plants the night before your trip." | The night-before mistake | Face to camera, holding a drooping plant |
| 3–12s | "Most of us give everything a huge drink before we leave. For a lot of plants, sitting in wet soil for days is the problem." | Too much water, too long | Close-up of soggy soil, water pooling in the saucer |
| 12–25s | "Instead, water a couple of days early, empty the saucer, and move them out of direct sun so they dry out slower." | Water early · empty saucer · move from sun | Three quick cuts: watering can, tipping the saucer, carrying the pot away from the window |
| 25–30s | "Save this for your next trip, and tell me which plant always dies on you." | Save for your next trip | Face to camera, pointing down at the caption |
Caption: "The night-before watering mistake, and what to do instead before your next trip 🌿"
Hashtags: pick 4–6 that describe the topic and audience, not just the biggest generic tags. For this one: #houseplants #plantcare #planttips #travelhacks.
Filming tip: shoot all the face-to-camera lines in one take, then all the close-ups. Batching shots by type is faster than filming beat by beat.
Pre-film checklist
Before you hit record, check:
- [ ] The angle fits in one sentence.
- [ ] You wrote three hooks and picked one that matches the payoff.
- [ ] Every beat has voiceover, on-screen text, a shot, and a time.
- [ ] The payoff arrives before the halfway point of a longer video, not right at the end.
- [ ] There's exactly one CTA.
- [ ] Every claim is true and you can back it up.
The same script works for TikTok, Instagram Reels and YouTube Shorts. Adjust the caption and hashtags per platform.
Automating the first draft with TikTok Script Pro
Going from rough idea to angle, three hooks, four timed beats, three layers per beat, and a caption takes a while every time. My team built an AI agent for this called TikTok Script Pro. It runs in a chat on Capafy, an AI agent marketplace.
What you send: a topic, trend, product, or half-formed thought, pasted as text.
What you get back, according to the listing, is one script package:
- the angle, meaning the single sharpest thing the video will say,
- three first-3-second hook options, with the strongest marked RECOMMENDED and the reason it works,
- a script table in HOOK → BUILD → PAYOFF → CTA order, with timed beats that each have voiceover, on-screen text, and a visual or B-roll cue,
- a caption with 4–6 hashtags and one filming or posting tip.
The listing says it's meant for TikTok, Reels and Shorts. It writes the script. Filming and posting are still up to you, and you should check what it gives you the way you'd check any first draft: is every claim true, and does it sound like you?
Pricing on the listing is $2.99/day, $5.99/week or $19.99/month.
Wrap-up
- Turn the topic into one angle before writing anything else.
- Write three hooks, pick the one the payoff actually delivers.
- Use HOOK → BUILD → PAYOFF → CTA, with one CTA.
- Give every beat a time, a spoken line, on-screen text, and a shot.
Once every beat has those layers, filming is mostly reading lines and getting the shots.
Disclosure: my team built TikTok Script Pro and sells it on Capafy.
Top comments (0)