draft with ChatGPT or Claude, deliver with open prompter

prompt an llm for a script in markdown, save the file into your synced folder, and read it on camera. no copy-paste into a teleprompter app.

a lot of people draft scripts with ChatGPT or Claude now. the model writes a clean first pass, you tighten it, and then you have to get it onto your phone to read it. that last step is where most teleprompter apps make you paste, import, or upload. open prompter skips it. the file is the handoff.

the loop#

here's the whole thing.

that's it. no paste box. no import button. no "copy your text here" field. you write the file once and the app reads it where it lives.

why markdown is the right handoff#

llms write markdown natively. ask for a script and you get back headings, bold, bullets, the works, already formatted. that's the default output, not a setting you have to turn on.

open prompter reads markdown natively. headings, bold, italic, bullets, numbered lists, blockquotes all render on the prompter the way you'd expect.

so the two ends already speak the same language. the model writes markdown. the app reads markdown. the file in between is the entire integration. no converter, no plugin, no API key.

and because it's a plain .md, nothing is locked anywhere. the script is a file in your folder. you can open it on your mac, edit it in any text editor, hand it to a collaborator. it's yours.

we don't write the script. that's on purpose.#

open prompter does not generate scripts in-app. there's no "write me a video" button inside the app, and there won't be.

that's a deliberate choice, not a missing feature.

you already have the best script tools on the planet. ChatGPT and Claude are right there, in the browser you already have open, with the context you've already given them about your voice and your audience. trying to bolt a worse version of that into a teleprompter app would be silly.

so we stay in our lane. the script is yours, written wherever you want to write it, sitting in your file, in your folder. we read it. we don't write it.

this also keeps the privacy story clean. open prompter makes no network calls of its own. nothing about your script goes through us, because your script never touches us. you draft with the model you chose, you save to the folder you chose, and the app reads what's there.

prompting the model for a script that reads well#

a script you'll read out loud is not the same as a script you'll read on a page. the difference is mostly in how you prompt. a few things that help.

• ask for short spoken lines. tell the model to write the way people talk, in short sentences, not in dense paragraphs. you're going to say this, not skim it.
• ask for a strong first line. the opening line is the hook. tell the model to earn the first three seconds before anything else. you can read more on writing hooks that survive the first three seconds.
• ask for headings on each section. markdown headings render on the prompter, so they double as section markers while you read. "intro," "the point," "call to action." they give you visual anchors as the text scrolls.
• tell it to skip stage directions. you don't need [smile here] or [cut to b-roll] cluttering the lines you're reading. ask the model to leave them out.

on that last one, there's a safety net. open prompter strips visual-direction cues like [B-roll: ...] at display time anyway, along with YAML frontmatter, footnotes, tables, and code blocks. it does that without touching your file. so if the model leaves a stage direction in, you won't be reading it off the prompter. but cleaner output is still nicer to work with, so ask for it.

keep iterating, it just syncs#

the first draft is rarely the take. that's fine, because nothing here is locked into the app.

read the script once on the prompter and you'll hear what's clunky. go back to the model and ask for a tighter version. or open the file in your editor and fix the line yourself. save it.

the phone catches up. open prompter watches the folder for changes and shows a last-edited indicator so you know the new version landed. sync latency is whatever your service delivers, usually ten to sixty seconds on iCloud. the app is honest about that. it doesn't fake instant.

you can also edit right on the phone. tap the pencil and change the file in-app. those edits save back through the system file coordinator, so iCloud syncs them to your mac like any other change. no proprietary container, no import-to-edit round trip. the file is always just a file.

so the loop tightens itself. draft, read, hear the rough spot, fix it in the model or the editor, save, read again. the script gets better and the file never leaves your hands.

your script, your file, your folder#

the reason this works is the same reason the whole app works. open prompter treats the file as the source of truth.

most of the category does the opposite. paste your script into their box. upload it to their cloud. trust them with the thing you wrote. some want a subscription before they'll even mirror the text for a beam splitter. your words end up living in someone else's product.

open prompter reads the markdown you already wrote, on the phone you already own, free and MIT-licensed forever. the model you like writes the draft. your folder holds the file. the app reads it. nobody in that chain owns your script except you.

next step#

open ChatGPT or Claude, ask for a short script in markdown, and save it as a .md in your synced folder. open open prompter, watch the script show up in your library, and read your first line.

if it needs work, fix it in the model or in the editor, save, and the phone has the new version a minute later.