The most important thing to know about Descript in 2026 isn’t a feature. It’s that the numbers on the pricing page stayed the same while the thing they measure changed underneath them.
Hobbyist still says ten hours. Creator still says thirty. Business still says forty. Those used to be transcription hours β you paid for the audio you asked Descript to transcribe. Since the overhaul on 23 September 2025, they are media hours, and a media hour counts every file you upload or record into the editor, whether you transcribe it or not. Identical numerals, different product.
That single substitution explains almost every complaint, every confused comparison, and every review published before late 2025 that’s still circulating as current advice.
What actually changed
Two new meters replaced one old one, and both can stop you independently:
- Media minutes β drawn down by anything you bring into the editor. Uploads, recordings, B-roll you audition and cut, takes you delete. Still images reportedly count as one second each.
- AI credits β a separate monthly pool consumed by Underlord actions, Studio Sound, Overdub, Eye Contact, green screen, dubbing and generative media. Before the overhaul these features were effectively uncapped inside your tier.
Reported credit costs give a feel for the burn rate: Studio Sound around 10 credits per use, Eye Contact around 10, dubbing around 15 per minute, text-to-speech around 5 per minute. Neither pool rolls over month to month. Purchased top-ups behave differently β they’re shared at the drive level and reportedly carry for up to twelve months β which is a strange asymmetry worth knowing: the allowance you pay for monthly expires, the extra you buy separately doesn’t.
Existing subscribers were moved onto Legacy/Sunset plans pending auto-migration with notice. Seat prices, incidentally, had been stable at roughly $0/$12/$24 from early 2022 through mid-2024; the jump to $16/$24/$50 arrived with the same September 2025 rework. So the price rose and the unit got broader in one move.
You now pay for the raw material, not the output
This is the part worth sitting with, because it cuts against what Descript is for.
The pitch β the genuinely great pitch β is that editing video should feel like editing a document. Record loosely, get a transcript, delete the bad sentences, rearrange, done. That workflow is built on bringing in more material than you’ll use. Three takes of the intro. Twenty minutes of interview for a four-minute cut. B-roll you try and discard.
Under transcription-minute billing, that was free β you only paid for what you asked it to transcribe. Under media-minute billing, every second of it counts against your quota before you’ve made a single decision. The tool whose whole value is cheap iteration now meters the iteration.
The practical discipline that falls out of this is faintly absurd, and it’s the honest advice: trim before you import. Cut the dead air, the false starts and the unusable takes in something else first, then bring the survivors into Descript. You’re doing rough triage outside a tool designed to make rough triage unnecessary β but it’s the difference between a Creator plan lasting the month and not.
Long-tenured users reporting higher bills without a visible change to their own quota line are usually experiencing exactly this. Their number didn’t move. Their unit did.
The 2026 plans
| Plan | Annual / monthly | Media | AI credits | Export |
|---|---|---|---|---|
| Free | $0 | 60 minutes/mo | 100, one-time | 720p, watermarked |
| Hobbyist | ~$16 / ~$24 | 10 hours/mo | 400/mo | 1080p, clean |
| Creator | ~$24 / ~$35 | 30 hours/mo | 800/mo | 4K |
| Business | ~$50 / ~$65 per seat | 40 hours/mo | 1,500/mo | 4K + Brand Studio |
| Enterprise | Custom | Custom | Custom | SSO/SCIM |
Annual billing reportedly saves somewhere between 23% and 33% depending on tier β the discount is steepest at the bottom, which quietly makes Hobbyist the plan most punished by monthly billing. Seats are billed per occupied seat, so a three-person team on Business is three subscriptions.
Two honesty notes. First, some 2026 reviews still describe these tiers in transcription hours; both they and the media-hour descriptions are quoting real Descript pages, just from different eras. Check which unit your own plan page names before trusting any comparison, and the help centre is the source of truth on how minutes and credits are counted. Second, allowances are the same globally while the amount charged varies by currency, tax and region β as does the pricing history, which a dated snapshot tracker shows more clearly than any current page can.
One small thing that recurs in reviews: upgrading mid-cycle reportedly doesn’t credit the time you already paid for. Minor in absolute terms, irritating in context.
What it’s genuinely excellent at
None of the above makes it a bad product. Descript holds around 4.6 out of 5 across two thousand-plus G2 reviews, and that’s earned.
Transcript-based editing for spoken word. For interviews, podcasts, talking-head video, tutorials and course material, it’s the fastest route from raw recording to publishable cut that exists. If you can edit a document, you can edit a video, and most people produce something usable within the first session.
Studio Sound. The most consistently praised individual feature β it makes mediocre recordings sound deliberate. On its own it justifies a subscription for anyone recording in an untreated room.
Rough-cut assembly for professional workflows. Underrated: Descript exports timelines to Premiere Pro, DaVinci Resolve and Final Cut. Cut the story in the transcript, finish in your real NLE. That’s a legitimate professional use, not a beginner shortcut.
Underlord. The agentic co-editor introduced in 2025 takes natural-language instructions β remove filler words, tighten pacing, cut three social clips β and executes multi-step edits. In 2026 it’s also reachable through a public API in open beta. It’s real, and it’s also the fastest way to spend your credit pool, so treat it as a tool with a meter rather than a background assistant.
The generative side β video via third-party models, images, stock avatars, dubbing across 30-plus languages β is capable but not why anyone buys Descript. If generation is your main need, a dedicated generative tool gives more per credit, and voice work is stronger in purpose-built voice platforms.
Who should buy which plan
- Free β genuinely usable for learning the editor on one short project. The watermark rules out anything client-facing, and 60 minutes goes fast once you count the material you discard.
- Hobbyist β a fortnightly podcast or occasional video, where ten media hours is comfortable and you don’t need 4K or the full AI suite.
- Creator β the real entry point for anyone publishing weekly. Thirty media hours absorbs normal over-recording, and 4K plus full AI access is where the tool stops feeling constrained.
- Business β teams, brand consistency, translation and dubbing at volume. Only worth it for the collaboration and localisation features, not for the extra ten media hours.
Don’t buy it for short-form vertical social video β CapCut is the right tool and free at the level most creators need. Don’t buy it for cinematic or heavily graded work; that’s a Resolve or Premiere job. And don’t buy it purely for transcripts β you’d be paying for an editor you won’t open, when dedicated notetakers meter that work more cheaply. The positioning question against timeline editors is laid out in more detail in our Descript versus CapCut comparison.
It’s also Windows and Mac only β no mobile app β which matters if you edit on the move.
The verdict
Descript remains the best tool in existence for turning recorded conversation into published video, and Studio Sound alone earns its keep for a lot of people. What changed is the shape of the bill, not the quality of the editor.
Budget on the assumption that you’ll consume roughly three to four times as much media as you publish, pick the tier that absorbs that, and trim before you import. Do that and it’s excellent value. Ignore it and you’ll hit a wall mid-project with a quota number that looks identical to the one you had last year. If you’re building a paid service around this kind of work, the business side is worth thinking through separately β this guide to becoming a freelance video editor covers pricing your time rather than your software. For the wider question of which subscriptions earn their place, see free versus paid AI tools and what still pays for content creators.
Frequently asked questions
What are Descript media minutes?
Media minutes count every file you upload or record into the editor, not just what you transcribe. Since the September 2025 overhaul they replaced the old transcription-minute quota, so takes you delete, B-roll you audition and re-records all draw down your allowance. Still images reportedly count as one second each.
Why did my Descript bill go up when my plan didn’t change?
Because the unit changed rather than the number. Ten hours used to mean ten hours of transcription; it now means ten hours of anything imported. If you record loosely and cut heavily β which is how Descript is designed to be used β you consume the same quota far faster than before.
How much does Descript cost in 2026?
Roughly $16 a month for Hobbyist, $24 for Creator and $50 per seat for Business on annual billing, with monthly billing around $24, $35 and $65. There’s a free plan with 60 media minutes and 100 one-time AI credits. Confirm the live figures on Descript’s own pricing page, since tiers have changed repeatedly.
Do Descript media minutes and AI credits roll over?
Monthly allowances do not roll over. Purchased top-ups behave differently β they’re reportedly shared at the drive level and carry for up to twelve months. That asymmetry means the extra capacity you buy outlasts the capacity included in your subscription.
Which Descript plan should I get?
Creator for anyone publishing weekly β thirty media hours absorbs normal over-recording, and it unlocks 4K and the full AI suite. Hobbyist suits occasional or fortnightly output. Business is worth it for team collaboration, brand controls and multi-language dubbing, not for the extra media hours.
Is Descript worth it in 2026?
For spoken-word video and podcasts, yes β it’s still the fastest path from raw recording to finished cut, and Studio Sound alone justifies the cost for many users. It’s the wrong tool for short-form vertical video, cinematic editing, or transcripts alone.
4 comments