Let an AI assistant run your teleprompter.

Count the jobs in a solo recording session. You are the writer, then the editor who cuts the writing down to something speakable, then the person who formats it for a prompter, then the operator who loads it, then the talent, then the operator again every time a take goes wrong. Only one of those jobs needs your face. The others are typing.

The work around the take

The script rarely starts as a script. It starts as a post you already published, a voice note, a page of bullet points, or an idea you had in the shower. Before it is readable off a prompter it needs to become short lines, broken by meaning, in sections you can restart from. That rewrite is real work, and it is the work most people skip, which is why so many prompter takes sound like someone reading a document.

Then comes the handling. Open the app, paste, split into sections, load, check the first line, start. After a stumble in the middle of section four, scroll back, find the start of section four, start again. None of it is hard. All of it pulls you out of the state you need to be in on camera, and on a bad night it is most of the evening.

Now you can just ask

AI assistants are already good at the rewrite. Ask one to turn a post into a spoken script with short chapters and you get something usable in seconds. The problem was the last step: the assistant lived in a chat window, the prompter lived in another app, and you were the clipboard between them.

That step is gone. Mirrento can now take instructions from the AI assistant you already use. You ask in plain words, and it does the handling for you. The connection between the assistant and Mirrento stays on your Mac. Mirrento itself still has no AI inside: the thinking is done by your assistant, and the prompter is simply something it can now reach.

What you can ask it

You talk to it the way you would talk to someone sitting next to you at the desk:

• "Turn this post into a script with short chapters." It writes the script and saves it in Mirrento. Your prompter keeps showing what it shows until you ask for the new one.
• "Make chapter two shorter." It edits the script. If that script is already on the prompter, the new text appears on the same chapter, so you stay where you were.
• "Load it." The script goes on the prompter, at the first chapter.
• "Go to chapter four." The prompter jumps there. "Next chapter" and "previous chapter" work too. No scrolling.
• "What is on the prompter right now?" It tells you which script and chapter are showing, and whether a take is recording.
• "How did that take go?" It reads the note Mirrento keeps for each take: how long it ran, your pace in each chapter, where you went back, where you paused, and the sentences that gave you the most trouble.

What it cannot do

This list matters more than the first one. It cannot delete a script or a recording. It cannot turn on the microphone, start voice-follow or record. It can put the words in front of you, and that is where its reach stops: you press record.

It also stays out of the way of a take in progress. It cannot put another script on the prompter while you are recording, and it cannot change the script you are recording: changes land between takes. And if you edit the same script in Mirrento at the same moment, the last save wins, so let one of you edit at a time.

On privacy, Mirrento records the sound of each take, not the picture: you film with your own camera or recording app. The note it keeps for each take holds positions and timings, never the words you said. To know where you are in the script, Mirrento follows your voice with the speech recognition built into your Mac, and it throws the words away when the take ends. The audio of the take stays on your Mac, and the assistant cannot get it.

A session that works

Here is a flow that keeps you in front of the camera and the assistant on the keyboard:

1. Hand over the source. Paste the post or the notes and ask for a spoken script in short chapters, one idea per chapter. Ask it to apply the rules of formatting a script for a prompter: short lines, no parentheses, numbers written the way you say them.
2. Read it once, out loud. Say what sounds wrong in your own words ("chapter two is too long", "I would never say that") and let the assistant edit. The edit lands on the prompter if the script is already loaded.
3. Load it and record. Ask it to load the script. Start the take yourself. After a stumble, ask for chapter four instead of scrolling for it.
4. Review the take. Ask how the take went, which chapter dragged and where you went back. Then ask for that chapter rewritten, and record it again.

Step four is the one people do not expect. The note does not hold your words, but it knows where you slowed down, paused and restarted, and that is usually enough to point at the sentence that needs rewriting.

Even better: an assistant that already sounds like you

How good the first draft is depends on how well the assistant knows the way you talk. A generic assistant writes a generic script. You spend the session rewriting it into your own voice, and you do it again next week, because it forgot.

Klyr is a clone of you. It starts from the identity you write yourself and keeps a memory of your work from one session to the next: your decisions, your clients, the way you explain things, and whatever you give it, like your past posts. So the script it writes starts closer to how you talk. When you correct it, it keeps the correction: it learns something once and never asks twice.

It drives Mirrento the same way as any assistant. You just ask it: write the script, load it, go to chapter four, how did that take go. The difference is that the words on the prompter already sound like you. Klyr is its own product, separate from Mirrento. The Klyr side of this is written up on getklyr.ai: how a clone that knows your voice turns an idea into a recorded video. If you want a clone of your own, start with the two ways to get one.

Setting it up

Assistants connect to apps through MCP, a standard way for an AI assistant to get tools it can use. Mirrento has it built in. It has to be open while the assistant works with it. Editing scripts also needs an active licence or the 7-day trial, the same as editing in the app.

If your assistant is Claude Code, paste this line in a terminal, once:

claude mcp add --scope user mirrento -- /Applications/Mirrento.app/Contents/MacOS/Mirrento mcp

The --scope user part makes Mirrento available to the assistant in every folder you work in, not just the current one. If you use the Claude Desktop app, the same thing goes in its configuration file as a server named mirrento, with /Applications/Mirrento.app/Contents/MacOS/Mirrento as the command and mcp as its argument.

Then ask the assistant to list your scripts. If it answers with your script list, it is connected.