SOCIAL

Short-form video script with shot notes

Use when you need a 30 to 60 second vertical video script that a non-actor can film in one take.

short-form-video-script.md

Download .md

You are writing a vertical short-form video script for {{PLATFORM}} on the topic below. It will be filmed by one person, talking to camera, with simple B-roll and on-screen text.

Topic and source material:
"""
{{SOURCE_MATERIAL}}
"""
Audience: {{AUDIENCE}}
Target length: {{SECONDS}} seconds.

Output a table with columns: Time | Spoken line | On-screen text | Shot note.

Rules:
- The first 2 seconds must show a result, an object, or a written claim. No "hi guys", no name, no channel intro.
- Spoken lines must be sayable in one breath. Under 14 words each.
- On-screen text is never a transcript of the spoken line; it adds or contrasts.
- Include one moment of visible proof, such as a screen, a document, or a physical object. If the source has no proof to show, say so and suggest what to capture.
- End on a specific next action, not "follow for more".
- Estimate total spoken time at roughly 2.5 words per second and state your estimate under the table. If it overruns, cut lines and show the cut version.

Do not write stage directions the filmer cannot do alone. No em dashes. No invented statistics.

Finish with three alternative first lines, ranked by how well they work with the sound off.

Fill in before running

Replace each placeholder with your own detail. The more specific you are, the less the model invents.

{{PLATFORM}}
{{SOURCE_MATERIAL}}
{{AUDIENCE}}
{{SECONDS}}

Getting a better result

Sound-off ranking matters more than the script - most viewers see the text before they hear you.
If the shot notes call for anything you cannot film today, ask for a version constrained to one location and a laptop.
Read the spoken lines aloud once; anything you stumble on gets rewritten, not the model's job to catch.

Questions about this prompt

When do I need this rather than a general script request?

When one person has to film it alone, today, on a phone. A general request returns lines nobody can shoot: two locations, a second pair of hands, a graphic that does not exist. This writes shot notes constrained to what a single non-actor can do in one take, and refuses directions the filmer cannot follow.

What do I need before running it?

Source material with something visible in it, the platform, the audience and a target in {{SECONDS}}. The prompt requires one moment of visible proof, so know in advance whether you have a screen, a document or an object to point at. If you have none, it will say so and suggest what to capture.

What comes back, and which part matters most?

A table of time, spoken line, on-screen text and shot note, a spoken-time estimate at roughly 2.5 words per second, and three alternative first lines ranked for how they work with the sound off. That ranking matters more than the script, because most viewers read the text before they hear anything.

What is the mistake?

Filming the spoken lines as written without reading them aloud once. The fourteen word cap is a rule about length, not about your mouth, and the line you stumble on is the one to rewrite. The word-rate estimate is an estimate too, so time your own read before you trust the total.