Getting Started with Descript

Edit audio and video by editing the transcript — powerful, intuitive, and easier to misuse than it looks.

Current prices: Free, from $24/person/mo monthly or $16 annual · See all plans

Last reviewed 2026-10-14 · for Descript

How it works

Descript transcribes your recording and then lets you edit the media by editing the text. Delete a sentence from the transcript and it disappears from the audio and video; rearrange a paragraph and the footage follows. For anything dialogue-driven, that collapses a timeline-editing skill into a word-processing one. You play it back, cut more, play it again.

What it is

An audio and video editor built around transcript editing rather than a timeline.

Aimed at spoken-word content — podcasts, interviews, tutorials — rather than at cinematic video work.

Metered on two separate things at once: how much media you bring in, and how much AI you use on it.

Distinct from voice-generation tools: this edits recordings of real people rather than synthesising new speech, though it can do both.

Before you start

The free plan gives a small monthly allowance of media, a one-time grant of AI credits, watermarked exports at 720p, and a single seat. It is a real trial, not a demo.

Paid plans step up from Hobbyist through Creator to Business, all priced per person; see current plans on the tool page. Monthly billing costs more than annual.

Media minutes and AI credits are different meters. Manual transcript cuts barely touch credits; leaning on the AI assistant exhausts them first. Which limit you hit depends on how you work, not how much you work.

This is a desktop application. There is no meaningful mobile editing.

Cloning a voice requires a recorded consent statement and a substantial audio sample, by design.

Security & privacy

Voice cloning here is gated: creating a voice model requires recording a spoken consent statement plus a sizeable sample, which is a deliberate barrier against cloning someone who has not agreed. That gate is real and worth respecting rather than working around.

The more distinctive risk is one no consent gate addresses. Transcript editing makes misrepresentation trivially easy. Deleting one word from a guest's sentence — a "not", a qualifier, a hedge — changes what they said, and the edit is seamless and invisible in the finished audio. Nothing warns you, because the tool cannot tell a tightening edit from a distorting one.

Working with other people's recordings

Get consent to record before you record, and be explicit that the recording will be edited.
Draw your own line between removing filler and changing meaning, and hold it when a cut would be convenient.
Send the edit to the person for review when their words carry any weight — a claim, a commitment, an opinion attributed to them.
Treat voice cloning of anyone but yourself as needing express, specific permission, whatever the tool permits.

Recordings are processed in the cloud, so interviews containing confidential or personal material are leaving your machine. Worth deciding before you upload, not after.

What it's good (and not good) at

Best at

Cutting dialogue-driven audio and video quickly
Removing filler words and false starts in bulk
Producing transcripts, captions, and short clips from one recording
Making editing approachable for people who have never opened a timeline editor

Okay at

Multi-camera and heavily layered video, which it handles but is not built for
Music and sound design, which are outside its focus
Collaborative editing, which works but is not its strength

Avoid

Colour grading, compositing, and finishing work — use a real video editor
Editing anyone else's words without a clear line on what you will and will not change
Heavy AI use on a small plan, where credits disappear faster than you expect

First things to try

Edit something by deleting words

Imports a short recording, waits for the transcript, then selects a few words in the text and deletes them — and plays the result back to hear them gone from the audio.

Why this one: The core idea sounds abstract and becomes obvious the moment you do it once.

Tweak: Use a recording of yourself. The first edits are always clumsy and you will not mind.

Remove filler words in one pass

Finds and strips hesitations across the whole recording at once.

Why this one: The single biggest time saving the tool offers, and the clearest demonstration of why transcript editing works.

Tweak: Listen back afterwards. Removing every hesitation can make speech sound unnaturally clipped — put a few back.

Watch your two meters for one project

Checks the media-minutes and AI-credits balances in your account before and after one project, to see which one moved more.

Why this one: Discovering which limit you hit first tells you which plan you actually need, and prevents the bill surprise this platform is known for.

Tweak: Do one project mostly manually and one mostly with the AI assistant. The difference in credit use is stark.

Set your own editing line before you edit a guest

Decides, before opening a guest's transcript, which cuts you will make to their words — fillers, false starts — and which you will not.

Why this one: Deciding what you will not change is much easier before you are staring at a cut that would make the episode better.

Tweak: Write it down. "Filler and false starts yes, meaning no" is a rule you can actually apply under time pressure.

Common pitfalls

Two meters, not one. Media minutes and AI credits run out independently. Heavy AI users hit credits first; heavy importers hit minutes. Size your plan by which one you will exhaust.
Seamless edits can change meaning. Cutting one qualifier from someone's sentence is invisible in the output and misrepresents them. The tool will not warn you, so the judgement is entirely yours.
Bill shock is the most common complaint. Costs climb fast once a team exceeds included limits. Watch the meters during your first month rather than at the end of it.
It is not a full video editor. Colour work, compositing, and advanced effects belong elsewhere. Descript is for cutting dialogue, and it is excellent at that.
Everything uploads. Recordings are processed in the cloud. If an interview contains confidential material, that decision needs making before the upload, not after.

Where to next

Voice and speech tools Video creation and editing

Want the full picture first? See the Descript overview for pricing, the free tier and related tools.