Overview
Descript transcribes video and audio, then lets you edit the media by editing the transcript — delete a word and the clip is cut. It also regenerates speech from text for voice-over fixes. Media hours and AI credits are metered separately, which is the detail that decides the right plan.
The paradigm shift: video becomes a document
Every video editor is built around the timeline: you drag clips, cut on the frame, align audio by eye. Descript inverts the model — it transcribes your footage, and you edit the text. Delete a word in the transcript and the video is cut at exactly that point. The transcript is the timeline.
That is a small change in interface and a large change in who can edit video: someone who has never touched a timeline can cut a podcast or a talking-head video by editing a document. People who edit professionally describe the same feeling — the mechanical work of cutting pauses, hesitations and dead air collapses into text deletion, and the hardest part of editing (judging a cut by eye and ear) becomes a reading task instead.
The workflow that makes it worth the subscription
The value concentrates in dialogue-heavy content, and the workflow is distinctive:
- Cutting the dead air. A podcast with a guest saying "um" forty times: find each one in the transcript, delete, and the cuts happen automatically — the filler-word count falls to zero and the runtime drops visibly, without touching the timeline.
- Fixing speech without re-recording. A talking-head video where a sentence is wrong: edit the sentence, and Descript's voice regeneration re-records the audio to match the new text — no re-recording session needed.
- Text as the editing surface. Search, replace and reorder operate on words, which is far more precise than scrubbing a timeline for the moment someone said something.
Professional users consistently name Descript the fastest route for dialogue editing, while keeping timeline tools like Premiere for anything that needs frame-level control. The limitation is the same one as any text-based tool: the regenerated voice is close but not identical, so word-perfect lip-sync and subtle emotional delivery still get a manual pass.
Two meters, and why it matters
Descript runs two separate meters, and the confusion between them is the pricing trap:
- Media hours — the amount of transcription and editing you can do, metered monthly (60 minutes free, 10 hours on Hobbyist, 30 on Creator).
- AI credits — a separate allowance for the generative features: voice regeneration, filler-word removal, studio-quality sound. The free tier grants 100 credits once, and higher plans come with larger monthly pools.
A workflow that is mostly cutting (an interview, a podcast) consumes media hours but little in AI credits. A workflow that leans on voice regeneration and filler removal burns credits quickly. The two-meter split means the right plan depends on which half of the work dominates — and teams that ignore the split find themselves buying more hours when their actual bottleneck was credits, or vice versa. The check before subscribing: count a typical project's split, because the plan that looks right on media hours can run dry on credits.
| Free | Hobbyist | Creator | |
|---|---|---|---|
| Media hours / month | 60 min | 10 hrs | 30 hrs |
| AI credits | 100 once | 400 / month | 600+ / month |
| Voice regeneration | Limited | Yes | Yes |
Where it fits
- ✓ Works for: podcasters and dialogue-heavy creators who cut interviews, where transcript editing is the fastest path; anyone who needs to fix a spoken mistake without re-recording; teams producing regular talking-head content who want non-editors to be able to cut; video professionals who want a fast dialogue pass before finishing in a timeline editor.
- ✗ Not a fit for: projects needing frame-level creative editing, complex effects or precise motion control, where timeline editors remain the right tool; content that is mostly visuals with little dialogue, where the transcript advantage disappears; users who generate heavily with AI features on the free tier, which runs out of credits long before running out of patience.