What it is and why
Descript flips editing on its head: it transcribes video and audio into text, and you edit the media like a document — delete a word in the text and it's cut out of the video. Automatic removal of "um"s and pauses, subtitles, and even an MCP integration (you can drive it from an AI agent). Reach for it for podcasts, talking-head videos and tutorials — anywhere editing means cleaning up speech rather than frame-by-frame magic.
How to do it (step by step)
- Sign up at descript.com — there's a free plan (with a transcription hour limit). Billing is in foreign currency.
- Create a project and upload your video or audio. Descript transcribes it automatically (note: Russian transcription is weaker than English — check it).
- Edit by text. Cut chunks by selecting and deleting words or sentences in the transcript — the video is trimmed in sync.
- Remove Filler Words. One button clears "um", "well", "like" and drawn-out pauses across the whole recording.
- Studio Sound — one-click audio cleanup (kill noise, level the volume).
- Subtitles. Auto-generate captions from the transcript, style them, export them burned in or as a file.
- Overdub / voice edits (where available) — re-record a mistake by typing the corrected text (you need consent for the voice!).
- Export in the format and aspect ratio you need. Via MCP you can automate the routine steps from an AI agent.
Ready-made recipe
A podcast in one evening: record the conversation → upload it to Descript → auto-transcript → delete the rambles and blown takes right in the text → Remove Filler Words clears the "um"s → Studio Sound levels the audio → generate the subtitles → export both audio (for podcast platforms) and a video version → hand the best moments to Opus Clip to cut into Shorts. Editing that used to take hours, done in an evening.
What it costs (honestly)
The logic: a free plan with a monthly transcription hour limit and basic features. Paid plans give you a package of transcription hours plus Studio Sound/Overdub and watermark-free export. Billing is in foreign currency. Exact figures and hour limits are on Descript's site and they change.
When it fits / when it doesn't
- Fits: podcasts, talking heads, tutorial videos; editing means cleaning up speech; you value speed; you need subtitles or a transcript; you want automation via MCP.
- Doesn't fit: you need precise cinematic editing (use a pro NLE); heavy projects with effects; flawless Russian transcription is critical.
Risks and responsibility
Voice cloning and voice editing (Overdub) is only allowed with the consent of the voice's owner (both ethics and voice rights). If you upload someone else's recordings, you need the rights. Transcripts containing personal data (names, details mentioned in conversation) deserve careful handling. If the video is an ad, ORD labeling applies on the publishing side. See the section disclaimer.
🤖 Prompt helper
You are a podcast producer working in Descript. I have a recording that is {format, length, language}. Give me: (1) the cleanup sequence (what to do in what order: transcript → filler words → Studio Sound → subtitles); (2) what to watch out for with Russian transcription; (3) how to turn the finished episode into raw material for Shorts clips; (4) a pre-export checklist. Step by step.