Descript

Delete a sentence from the transcript, and it's gone from the video too — genuinely one of the smartest editing paradigms in the category, now billed differently than it was in 2025.

Video · Video Editing, Subtitles & Captions · 4.4 ★

What is Descript?

Descript's founding idea has aged extremely well: rather than scrubbing through a timeline to find and cut a moment, you edit a text transcript of your recording, and deleting a word or sentence removes the corresponding footage automatically. For podcasters, YouTubers, course creators, and anyone producing spoken-word content, that paradigm genuinely reduces editing time by an estimated 60 to 70%, and it's the reason more than 6 million creators use the platform. Underlord, its AI co-editor, goes further than a simple assistant — it's agentic, meaning it can take initiative on a request rather than just offering a one-off suggestion, drafting scripts, summarizing recordings, and identifying viral-worthy clips automatically. Descript's own help documentation is refreshingly candid about where it still falls short, noting plainly that Underlord "might overpromise, make incorrect assumptions, or follow you into workflows it's not equipped to complete" — an honest admission you don't see from every AI product.

Beyond text editing, Overdub 3.0 clones a voice from just three minutes of training audio, with a newer version adding emotional intonation control for whispering or excited delivery; Studio Sound 4.0 automatically separates voice, background, and music stems while removing reverb; Eye Contact AI corrects on-camera gaze, now working with glasses and extreme angles; and Generative B-Roll can produce context-aware footage for a highlighted sentence using integrated diffusion models. The genuinely important thing to know before subscribing is a September 2025 pricing overhaul that changed the entire cost structure: Descript moved from billing by transcription minutes to a broader "media minute" model that counts any audio or video brought into the editor, whether transcribed or not, and introduced metered AI credit top-ups for features that used to be marketed as unlimited. One independent analysis put it plainly — the same workflow that cost a podcaster around $30 a month in early 2025 can run noticeably higher in 2026 once real credit consumption is factored in, and the free tier was deliberately tightened during that same overhaul specifically to push more users toward a paid plan.

⚡
60-70%
Estimated editing time reduction for spoken-word content
👥
6M+
Creators using Descript
📅
Sep 2025
Pricing overhaul — moved from transcription-minute to media-minute + credits
🎙️
3 minutes
Training audio needed to clone a voice with Overdub 3.0

Editing-time estimates and the September 2025 pricing overhaul details are drawn from multiple independent 2026 reviews (Sonix, Hack'celeration, max-productive.ai) and Descript's own published documentation. Underlord's self-described limitations are quoted directly from Descript's help center.

Key features

📝

Text-based editing

Edit the transcript, and the corresponding audio and video are cut automatically.

🤖

Underlord AI co-editor

An agentic assistant that can draft scripts, summarize recordings, and act on editing requests directly.

🎙️

Overdub voice cloning

Clone your own voice from just three minutes of audio, now with emotional intonation control.

🔊

Studio Sound audio restoration

Automatically separates voice, background and music while removing reverb.

👁️

Eye Contact AI

Corrects on-camera gaze so a speaker appears to look directly at the lens.

🎬

Generative B-Roll

Generates context-aware supporting footage for a highlighted line in the transcript.

Pricing

Media minutes now count all audio and video brought into the editor, not just what's transcribed — a real change from the pre-September 2025 model. AI credit top-ups are sold separately when you exceed your plan's pool and expire 12 months after purchase. Descript is SOC 2 Type II compliant. Prices reflect Descript's published pricing as of July 2026; lower figures reflect annual billing.

Available models

Underlord Descript's agentic AI co-editor, with selectable underlying models on paid plans
Overdub 3.0 Voice cloning model, now with emotional intonation control

Integrations & platforms

Desktop app (Mac, Windows) Web app Lyrebird API (Enterprise)

No offline editing is currently supported — an internet connection is required.

Pros, cons & best for

👍

Pros

  • Text-based editing genuinely transforms spoken-word editing speed
  • Broad, well-integrated AI toolset covering voice, audio and B-roll
  • Refreshingly honest about Underlord's current limitations
👎

Cons

  • September 2025 overhaul made real costs less predictable than before
  • AI credits can run out mid-month on heavier workflows
  • No offline editing, and stability on complex projects is still maturing
🎯

Best for

  • Podcasters, YouTubers and course creators editing spoken-word content
  • Teams wanting voice cloning, dubbing and audio restoration in one tool
  • Not the pick if you need offline editing or perfectly predictable monthly costs

Take a look inside

Our verdict

4.4 / 5

Descript's core insight — that editing spoken-word video should feel like editing a document, not scrubbing a timeline — remains one of the genuinely smartest ideas in this whole category, and the 60-70% editing time reduction it delivers for the right kind of content is real, not marketing. The breadth of what's bundled around that core idea, from voice cloning to automatic audio restoration to generative B-roll, makes it a legitimate one-stop platform for podcasters and creators rather than just a transcription tool with extras. The September 2025 shift to media-minute billing and metered AI credits is the real thing to plan around: your actual monthly cost now depends more on how you use the AI features than the plan price alone suggests, so it's worth testing your real workflow against the free tier before committing to a paid plan.

FAQ

Is Descript free to use?

Yes, a free plan includes 60 media minutes a month, though exports are watermarked and AI features are capped — it was deliberately tightened during a September 2025 pricing overhaul to encourage upgrades.

What changed with Descript's pricing in September 2025?

Descript moved from billing by transcription minutes to a broader "media minute" model that counts any audio or video brought into the editor, and introduced metered AI credit top-ups for features previously marketed as unlimited — a real shift that's made monthly costs less predictable for heavy AI users.

What is Underlord?

Descript's agentic AI co-editor, which can draft scripts, summarize recordings, and act directly on editing requests rather than only offering suggestions — Descript's own documentation candidly notes it can still overpromise or misunderstand a request at times.

How does text-based editing actually work?

Descript automatically transcribes your recording into an editable text document; deleting a word or sentence from that transcript removes the corresponding audio and video footage automatically.

Can Descript clone my voice?

Yes, Overdub 3.0 creates a digital voice clone from just three minutes of training audio, with a newer version adding emotional intonation control for different delivery styles.

Does Descript work offline?

No, an internet connection is required — there's currently no offline editing mode, which is worth knowing if you frequently work without reliable connectivity.