How to format scripts for Supershoot

The format

Give each script a title line, put the words you say in short paragraphs, and put everything else in [square brackets] on its own line. That is the whole format.

Script 1: The 5 AM Secret

Ever wonder why our croissants are gone by nine?

[hold up a croissant]
We start baking at five, every single morning. The butter is cold and the dough is rested.

Come early, or call ahead and we will save you a box at [Bakery Name].

Script 2: Why We Use Real Butter

[SPOT A]
Anyone can say butter. Ours is cold and folded in by hand.

[SPOT B]
Cheaper fat melts early. Real butter makes the layers.

Titles

Script 1: Title is the clearest. Script 1, Video 2: Title, 1. Title, #3 Title, ## Title and a bold or ALL CAPS title line also work. A script with no title becomes "Script N". Leave a blank line before each title. A line of --- between scripts also works.

Spoken words and beats

The importer groups your spoken sentences into beats of two or three. It starts a new beat at a paragraph break once a beat has two sentences. You can edit the split in the preview before you create the shoot.

Directions

Anything in [square brackets], or a whole line in (parentheses), shows in the prompter and never gets read aloud. Crew notes start with a label such as On screen:, B-roll:, Shot:, Music: or Visual: and sit above the reading flow. A direction can sit inside a line too: [smile].

Camera spots

Optional. Put [SPOT A], [SPOT B] on their own line before the lines you shoot from that spot. These forms read the same way: Spot A:, Camera A, Cam A (wide), [A], (A), A: (needs two or more spots) and SPOT 1A (location 1). With no markers, spots are assigned for you and rotate so the same spot never repeats back to back.

Placeholders

[Business Name], [address] and [price] stay in the sentence and show highlighted, so you fill them in on camera.

Left alone on purpose

Numbered list items keep their numbers. Emoji and curly quotes stay in the spoken text.

What the importer cleans up for you

Without the prompt, Supershoot already handles most of what AI tools wrap around a script. It went through a test set of 62 pastes: real Claude output, sample ChatGPT, Gemini, Copilot and Perplexity style output, Google Docs, Word and Apple Notes pastes, and deliberate junk. All 52 raw pastes import with the right script count and no chatter in the spoken text.

• The chat intro and sign-off, plus a trailing Tips section or Sources list.
• Labels like Hook, Body, CTA, Voiceover, Audio, Caption and Visual, in any dress: **Hook (0-3 sec):**, [CTA], a bare HOOK, Reveal (3-9s):.
• Quotation marks around spoken lines.
• Tables. The Audio column becomes the script, other columns become crew cues, a content-plan table goes to notes, and a Title column makes one script per row.
• Scene 1 lines. They drop out of the spoken text and never split a script. A paste of only Scene blocks imports as one script.
• Citation marks and Sources lists.
• Emoji headings, trailing duration notes and titles like ## 2. 3 Mistakes.
• Windows and old Mac line endings, and the odd breaks from Google Docs and Apple Notes. A header paragraph above the first title (Apple Notes) becomes the shoot header.
• A closing paragraph, only after the last script and only when it offers more help ("Would you like me to", "Let me know if"). A call to action aimed at viewers stays.

One thing it can't know: a plain explanation paragraph with no quoted script beside it looks like spoken text, so it stays. The prompt below asks the AI to leave that kind of paragraph out.

The formatting prompt

Paste this into an AI chat, then paste your scripts right after it, in the same message. It asks for your wording kept exactly and only the formatted scripts back. The app's Need help formatting? row copies this same text.

Reformat the scripts I give you so a teleprompter app can read them. Keep my wording exactly. Do not rewrite, shorten or add to it.

Rules:
1. Plain text only. No markdown, asterisks, tables, bullet points, hashtags or captions.
2. Start every script with a title line: Script 1: Short Title. Then Script 2: and so on. Leave a blank line between scripts.
3. Under each title, write only the words I say out loud to the camera, in short paragraphs with a blank line between them. Leave out background explanation, research notes, citation numbers and sources.
4. Put anything I do not say in [square brackets] on its own line: actions, camera moves, on-screen text, B-roll. Example: [walk toward camera]
5. Delete labels such as Hook, Body, CTA, Voiceover, Scene, Visual and Audio. Delete timestamps and the quotation marks around spoken lines.
6. If my scripts name camera spots or angles (A, B, C), put each marker on its own line before that part: [SPOT A], [SPOT B]. If they do not, add no markers.
7. Keep placeholders such as [Business Name] inside the sentence where they belong.

Example of the finished format:

Script 1: The 5 AM Secret

Ever wonder why our croissants are gone by nine?

[hold up a croissant]
We start baking at five, every single morning. The butter is cold and the dough is rested.

Come early, or call ahead and we will save you a box at [Bakery Name].

My scripts are pasted right after this message.

Reply with only the formatted scripts.

Where to paste it

Claude, ChatGPT, Gemini, Copilot, Perplexity or any other chat AI. It is plain text with no setup.

I tested it with Claude only. I ran it through Claude (Sonnet) on ten messy pastes, including four real Claude outputs and samples styled after ChatGPT, Gemini and Perplexity. Every answer imported with the right script count, no labels or chatter in the spoken text and the right camera spots. I haven't run it inside ChatGPT, Gemini, Copilot or Perplexity themselves, so if one of them adds extra chatter, Supershoot still strips most of it on import.

Import messages and what to do

When the importer can't read a paste, it says why under the text box. These are the exact messages. Where an AI can fix the problem, the message has a Copy prompt button inside it. An empty box shows nothing, and the Create button reads "Paste scripts to start". Quieter notes, like a chat intro that was skipped or a table that was read, show in the preview's warning list.