How to Edit a Podcast in Descript

Most “edit a podcast with AI” posts skip the job that actually decides the subscription. The episode is still a ramble. You need to delete the tangent at 18:40, clean the kitchen-mic guest, and hand your host an MP3 or an MP4 — not a folder of verticals you have not listened to. That cut happens in a transcript. This guide is the working path — file, project, composition or sequence, Studio Sound, filler review, export — using the one editor we send for this brief: Descript.

When Descript is the editor — and when you should clip instead

Use Descript when the words are the edit. A weekly interview, a two-host show, a solo essay you recorded in one take: you read the script, you delete, you move on. Studio Sound and Remove filler words sit on that text layer. Rooms will record the next session if you do not already have a booth export. That is the radio edit.

Skip Descript-as-clipper if the episode already exists and the bottleneck is Shorts, Reels, or TikTok. Finding 40-second hooks inside a finished hour is a clipper job. The working path on this site is How to make podcast clips with Klap — an ordinary guide link, not a second money hop. Skip both if you already know the 40 seconds and the job is effects or a template. That is CapCut — editorial on this site, no affiliate desk.

The seven-step podcast radio edit

1. 01

   Confirm the episode is yours to cut, then get a file — not an RSS dump

   This workflow is for a show you host, or a session you have written permission to edit. Guest likeness, a licensed music bed, and a sponsor read are not automatically yours to recut or republish. Get the edit and clip rights in the guest release. Then export a recording: a mixed WAV/MP3 from the booth, separate host and guest tracks from Riverside or a recorder, or a Rooms session you captured in Descript. An RSS XML feed is not an editor input.

2. 02

   Start a project for this episode and import the recording

   Treat one episode as one project. Descript’s own project docs describe a project as the binder and compositions as the pages inside — keep the binder small. Import from your computer, from Zoom if that is the source, or from a phone upload. You can also record straight into a composition, or remotely in Rooms (the rebranded SquadCast), which captures each participant on a separate track. Transcription starts after the file lands. Every uploaded or recorded file counts against media minutes — a one-hour interview plus a guest track plus B-roll is three files against the same pool.

3. 03

   Decide composition vs sequence before you start deleting sentences

   A composition is the document you radio-edit: delete words in the script and the media cuts. A sequence is the official way to keep multiple related files — host mic, guest mic, camera, lav — aligned as one script item. When you drag several files in, Descript prompts you to Combine into sequence or Keep separate. Combine when the tracks are the same conversation. Keep separate only if the files are sequential (intro, then interview, then outro) and should sit one after another. If a recorder started late, align tracks in the Sequence Editor before you cut from the script — a misaligned sequence makes a jumbled combined transcript.

4. 04

   Radio-edit the transcript: delete the ramble, keep the conversation

   Read the script. Delete the three-minute tangent, the “as I was saying off-mic,” the second take of the same answer. The media follows the text. That is the product. Use the timeline when you need a tighter clip boundary or a word-gap adjustment — Descript’s podcast guide still treats the script as the main surface and the timeline as the fine-tune. Do not ask Underlord to invent a show you have not cut. A prompt like “tighten the pacing” is a junior-editor pass you still watch.

5. 05

   Apply Studio Sound, then listen on headphones

   Studio Sound is the published cleanup: AI that reduces background noise, echo, and other distractions on spoken voice. Official help applies it from the Properties panel (Audio Effects) on a script layer, or on a single track inside the Sequence Editor, or by asking Underlord to enhance the audio. It is a file-level effect — once on, it follows that file everywhere it appears. It burns AI credits. Listen after it runs. A treated room usually wants less of it than a laptop mic in a kitchen. Do not stack it on a file you already ran through a different enhancer without checking for metallic artifacts.

6. 06

   Remove filler words — review the list, do not guess every um

   Descript detects common fillers such as “um,” “uh,” and “you know” and underlines them in the script. Official help opens Remove filler words from the AI Tools panel. You can delete the word and the audio, delete and replace with a gap clip, ignore (strike the text, remove the audio), or remove the word from the transcript only. Prefer the option that keeps the cut listenable. Enable Avoid harsh cuts when you want the tool to skip fillers that would clip into the next word. Do not bulk-delete every “like” on a guest who uses it as a beat. Watch the first two minutes after the pass. This also burns AI credits.

7. 07

   Export for YouTube and your RSS host — then clip the finished episode elsewhere

   Open Export. For a file you will upload yourself, choose Local export. Audio formats on the published audio-export page are .m4a, .wav, and .mp3; video local export is MP4. For a video episode, YouTube is a destination in the web app — Shorts are not supported on that path, so a vertical still wants a download. Descript does not generate your show’s RSS feed. Export the enclosure (usually MP3 or M4A, with metadata if you fill it in), then upload it to the host that already publishes your feed — Buzzsprout, Transistor, Libsyn, or whoever you pay. Some hosts accept a direct publish from Descript; confirm the live destination list in the export window. If you still need 5–15 verticals, that is a clipper job on the finished file — see the Klap podcast-clips how-to. Do not buy Create Clips hoping it replaces that batch.

One mixed file vs a sequence vs a clipper

“Multitrack podcast editing” is a search. In Descript the published objects are a composition (the page you edit) and a sequence (the container that keeps related tracks aligned). A clipper is a different product. Pricing on the Descript hop matches descript.com/pricing as of August 2026. Watch both meters — media minutes and AI credits — before you treat Hobbyist as a weekly-show plan.

Import the recording — computer, Zoom, phone, or Rooms

Open a project for this episode. Official getting-started copy gives you two starts: record in Descript, or import an existing file. Import paths they publish include upload from your computer, import from Zoom, and upload from a phone. Direct recording lands in the composition. Rooms records guests remotely and keeps each participant on a separate track — useful when you do not already pay Riverside for ISOs. Rooms hours are capped by plan; the Descript review has the August 2026 grid.

If you recorded in Riverside or a booth, export the session you will actually edit. A mixed program file is the right input for a composition. Separate tracks are the right input for a sequence. Do not hand the editor an RSS XML document from Buzzsprout or Transistor — that feed is how listeners subscribe, not how you cut. Download the enclosure only if you are recutting a published episode you already own.

Transcription is the next automatic step, not a button we need to invent. Wait for the script before you start deleting. Media minutes start when the file is in the project. A one-hour interview you also imported as a guest ISO and a camera card is three files. Plan the import before you drag everything in “just in case.”

Composition vs sequence — the actual multitrack split

Descript’s project overview is blunt: a project is a binder, compositions are pages, sequences hold multiple media tracks. Create a new project per episode so the binder stays small. Inside it, keep one composition for the radio edit you will export. Extra compositions are for an alternate cut or a social clip you already chose — they are not a substitute for a clipper on a two-hour interview.

A sequence is how you keep host and guest (or camera and lav) moving together. Official create-sequence help: drag multiple files into the script, label speakers, then Combine into sequence. Keep separate lays the files end to end. The sequence shows up in the script and timeline as one item; cuts apply to every track. Open the Sequence Editor when one track needs volume, Studio Sound, or a timing fix — those edits follow the sequence everywhere it appears. Official help allows up to 14 tracks in a sequence. If a recorder started late, align first. A combined transcript built on drifted tracks will read like two conversations at once.

You do not owe the product a sequence. A mixed MP3 from the booth is a composition job. Forcing ISOs into a sequence you will immediately flatten is how people burn media minutes and then wonder why the script looks duplicated. If two mics transcribed the same room, official Sequence Editor help lets you include or exclude a track from the combined script so the guest is not printed twice.

Studio Sound, then filler — both burn AI credits

Studio Sound is the cleanup name on the marketing site and in help. It reduces background noise, echo, and other distractions on spoken voice. Apply it from the Properties panel in a composition, or on a selected track in the Sequence Editor, or by asking Underlord to enhance the audio. Official help is explicit that it is file-level: one toggle follows every instance of that file. It wants a network connection. Listen on headphones after it runs. A treated booth often needs less than a default pass. A kitchen-mic guest is why the feature exists. Do not assume it will save a file that clipped in the recorder.

Filler is a separate pass. Official Remove filler words help detects “um,” “uh,” and “you know,” underlines them, and opens a review list from the AI Tools panel. We are not inventing a toolbar icon beyond that. Handle each hit the way the take needs: delete, delete-and-gap, ignore, or transcript only. Avoid harsh cuts is the published control that skips fillers that would chew the next word. That analysis takes a moment. Use it. Then watch the open. A host who says “um” before every proper noun will sound gated if you delete the lot. A guest who says “you know” as a shrug may want the shrug left in.

Underlord can chain these — remove filler, tighten pacing, add chapters — from a prompt. Full access starts at Creator on the August 2026 grid; Hobbyist is limited. Treat the result like a junior editor. Chapters are useful show-notes fodder; official Sound Good help can generate markers or a timestamped list. They are not a substitute for listening. Both Studio Sound and filler removal spend AI credits. The Free plan’s 100 credits are one-time. Hobbyist’s 400 go quickly once Underlord is in the loop — that complaint is consistent enough in the Descript review to plan around.

Export for YouTube, then for the RSS host

The episode is not done until a file leaves Descript. Official export overview: click Export, pick a destination. Local export writes a file to your computer. Integrated destinations publish to YouTube or to podcast hosts Descript lists in that window. We describe the action. We do not invent a host that is not in your dropdown.

For RSS, you want an audio enclosure your host will accept. The published audio-export page lists .m4a, .wav, and .mp3, plus channel, sample-rate, bitrate, and loudness options including −16 LUFS and −18 LUFS. Fill metadata — show title, episode title, description, artwork, chapters — when the format supports it. Official help says .wav does not carry artwork or chapter markers. Descript does not create the show’s RSS URL. Your host does, after you upload the enclosure. If a host you already pay appears as a destination, you can publish there from the same Export window. Confirm it. Then paste nothing into Apple Podcasts except the feed you already use.

For a video podcast, YouTube is a published destination in the web app: title, description, tags, privacy. Official caveats we will not sand off: Shorts are not supported on that publish path; brand and shared-access channels are not supported; thumbnails set in Descript apply to the Descript share page, not automatically to YouTube Studio. If the cut is a Short, download the vertical and upload it in Studio. Unverified YouTube accounts still hit the 15-minute upload cap — confirm the live Studio limit.

Free-plan exports are 720p with a watermark. Hobbyist unlocks 1080p. Creator and Business list 4K. Do not publish a watermarked demo to a feed you already monetize. Sit on a paid plan before the episode is the one subscribers get.

Which Descript plan covers a weekly show in 2026

Read the live page, not a remembered transcription-hour sticker. Descript replaced that model in late 2025 with two meters. Media minutes count every file you upload or record. AI credits burn on Studio Sound, filler-word removal, Underlord, and AI Speech. The public grid we can verify in August 2026 (descript.com/pricing): Free at $0 (60 media minutes/month, 100 one-time AI credits, 720p watermark, 5GB). Hobbyist at $16/month billed yearly ($24 month-to-month) for 10 media hours and 400 AI credits. Creator at $24/month yearly ($35 monthly) for 30 hours, 800 credits, 4K, and full Underlord. Business at $50/month yearly ($65 monthly). Unused allowances do not bank the way people hope. Confirm checkout.

For a weekly hour, Creator on annual billing is the realistic floor if you will actually run Studio Sound and filler removal. Hobbyist will teach you the transcript. It will not survive a month of Underlord. Do not buy Business for one English show unless you need the extra hours, Brand Studio, or the published 30+ language dubbing. Full side-by-side with the clipper you may add after export: Descript vs Klap. If you already finish verticals on a timeline: Descript vs CapCut.

When to hand the finished episode to a clipper

Export first. The clipper wants a picture or a finished program, not a half-cut composition. If you publish video, the YouTube URL or the mixed MP4 is the input. If you publish audio-only, build a talking-head or still-plus-waveform video before you mine Shorts — that constraint lives on How to make podcast clips with Klap. Descript’s Create Clips / Underlord can pull highlights. That is a side door. It is not a virality-scored batch with a clip-count meter. Use it for a couple of moments you already heard. Do not use it as the season factory.

Klap stays a foil on this page, not a Try button. CapCut stays a finish tool if a keeper still needs a template: capcut.com. Riverside and YouTube stay source or destination tools. Voiceover and clone questions — a host who wants a retake without a reshoot — are a different brief: ElevenLabs vs Murf vs Synthesys and How to clone your voice for YouTube. Descript’s own AI Speech Regenerate can fix a flubbed word in your voice; that is a patch, not a channel VO stack. None of those names are hops here.

We do not print an affiliate commission rate, and we do not invent a “this will earn you $X” line. The product question is whether the radio edit on your episode survives a headphone pass after Studio Sound and a filler review.