A script is a map, not a cage
The best talking-head videos are not read word for word. They are mostly the script, plus the small things that only happen live: a sentence said a better way, a quick aside, a joke that lands because it was not planned. Those moments are often what makes a video sound like a person.
A prompter that scrolls on a timer punishes every one of them. It does not know you stopped reading, so it keeps going. By the time you come back to the text, your line is gone and your eyes are hunting up the screen, on camera. After that happens twice, you stop improvising. That is how a prompter makes you sound like you are reading: not only through the pace, but by taking away your freedom to leave the page.
The text moves when your words match
With Voice follow on, Mirrento listens to you through the Mac's own speech recognition and compares what you say with the words in the script. The text moves only when what you say matches what is written. That one rule changes what happens when you leave the page:
- You stop to think. It waits. Nothing moves while you are silent.
- You add a sentence that is not in the script. It waits on the line you left. Your new words are not on the page, so it has no reason to move.
- You come back to the written words. It picks up where you are, as soon as what you say matches the text again.
- You skip a sentence or two. It catches up to where you landed.
- You go back a few words to say a phrase again. It follows you back.
- You jump to a part further away, ahead or behind. It moves there too, once it is sure. It waits for a few clear words in a row before it jumps, so a word that happens to appear elsewhere in the script does not throw it.
Finding your line when you come back
Going off script is only half the problem. Coming back is the other half, and it is where a timer prompter makes you search on camera. While Voice follow is on, the prompter is built for the return:
- The word you are on is highlighted. When your eyes come back, they land on it.
- What you have read is dimmed. Everything you have not read yet stays at full strength, so the next line is the clear one.
- The view moves in blocks, not a crawl. The lines in front of you stay still while you read them. The view only moves when you leave that block, so the text is where you left it when you look back.
When it loses you
It will sometimes. Most often after a long stretch off script, or a jump back to a part that sounds like another part. When that happens, it does not run away from you. It stays where it was, and you can put it back by hand in a second:
- Click a word on the prompter, and it carries on from that word.
- Up and down arrows, or the space bar, move by one block. Home goes back to the start.
- Left and right arrows move by chapter, from the prompter or the main window.
- From the chair, the phone remote moves the text by a block and follows you from there.
- Out loud, say the wake word and a command, such as "prompter, again" to go back to the start of the chapter. The full list is in you speak to your prompter.
Every one of these also tells Voice follow where you are, so it carries on listening from the new spot.
It still needs the words
To be fair about the limits: Voice follow follows the script. If you rewrite a whole section in your head and say it in completely different words, there is nothing to match, and the text will wait for you at the start of that section. That is the right behaviour, but it means an ad-lib works best as a detour, not a new route. Say your aside, then come back to a sentence that is on the page.
It also listens best in a quiet room, with one voice. Background noise and long silences make it harder. The speech recognition runs on your Mac, and for some languages macOS has to download the on-device model first; until then Mirrento asks you to add the language in System Settings instead of sending your voice anywhere.
If you have been cutting your best asides to keep up with the scroll, the 7-day trial is enough to record one script with a detour in every chapter and hear what the takes sound like when you keep them.
Write a script that leaves room
- Mark the spots where you want to talk freely. A short cue like
[[story]]or[[your own words]]reminds you that this is a detour, and it shows on the prompter as a small label, not as words to read. More on cues in how to format a script for a teleprompter. - Start the next part with a clear sentence. After a detour, the first written words are how the prompter finds you again. A distinct opening line makes that instant.
- Keep chapters short. If a detour goes wrong, going back to the start of the chapter costs a few seconds, not the whole take.
The script stays your safety net. The difference is that leaving it no longer costs you your place.