Thinking out loud
"So this is, um, where you'd, actually let me go back to Billing first…"
Polysh for agents is in early access. Your coding agent records the demo, and Polysh finishes it. Join the waitlist →
Polysh rewrites what you said into a clear script, reads it with an AI voice, and re-times your footage one sentence at a time. Nothing is re-recorded.
Free during early access.
The screens are right. The sound isn't, and re-recording means getting every click right again.
"So this is, um, where you'd, actually let me go back to Billing first…"
Six seconds of silence while the plan options load.
A laptop fan, a notification sound, and a mic that is too far away.
Polysh transcribes your speech and notes when each word was said.
AI turns it into clear sentences, keeping every step in the order you did it.
Preview an AI voice. It reads only the script you approved.
Each part of the footage is fitted to the sentence that describes it.
| Stage | What you get |
|---|---|
| Transcript | "OK so, Billing is, um, here, and you'd pick, uh, the yearly one I think it's called Annual…" |
| Script | "Open Billing. Choose Annual, then select Update plan." |
| Video | The click on Update plan lands as "Update plan" is spoken. |
Why the timing holds
New narration is rarely the same length as the old. Stretching the whole video to fit lets clicks drift away from their words.
The rewrite is a draft. Read it beside the video and change any line before it's voiced.
Professional, casual, or technical. Regenerate if it doesn't sound like your team.
Reword, split a sentence, or add the warning you forgot to say. Each line stays tied to its part of the video.
The team word list keeps product names and acronyms spelled and pronounced your way.
After an edit, only the narration around that sentence is voiced again.
Record with the Polysh recorder and each click becomes a zoom. Because the video is re-timed per sentence, the zoom arrives as the narrator names the button.
Turn auto zoom off if the screen is already easy to read.
Skip the zoom on a click that doesn't matter, right from the timeline.
No. The voice, trims, zooms, and timing are applied on export. The file you recorded or uploaded stays as it was.
Write the script yourself in the editor, and Polysh voices and times it.
Polysh is built to replace rough narration. If your audio is already good, the trims, zooms, and captions may be all you need.
No. Any MP4 works for the script, voice, sync, and captions. The recorder adds the click log that gives a zoom on each click.
You see an estimate first. Credits are charged only when the export succeeds and returned if it fails or you cancel. See how pricing works.
Know how long your narration will run before you record it.
Open toolMeasure your speaking pace in words per minute and see how it rates.
Open toolFind out whether your script fits the target length and what to change.
Open toolTurn an SRT subtitle file into a WebVTT file that HTML5 players accept.
Open toolRun your next walkthrough through Polysh. Founding teams keep their rate.