How to Use Descript to Edit a Company Podcast Episode

The short answer: load your product and guest names into the transcription glossary, record the episode in a Descript Room, name the speakers, cut the conversation by deleting text, run Remove filler words and Shorten word gaps, switch on Studio Sound, mark chapters, let an assistant draft the show notes while a person listens to every cut, and export an MP3 with the chapters inside.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

A company podcast rarely dies in the recording. It dies in the edit, because the edit means evenings of scrubbing a waveform and the marketing team has no spare evenings. Descript moves that work into a transcript, which anyone who has edited a document can handle on the first day. For plans and what each usage meter costs, see the Descript glossary entry.

Before you record:

• Enough Room time on the plan. Remote recording comes with every plan: Hobbyist includes 5 recording hours a month and Creator 15, and media hours are counted separately, per the pricing page (checked on 24 October 2026). Hobbyist is $24 a month on monthly billing.
• A laptop or desktop for every guest. Rooms is not supported on mobile, and Descript says it cannot troubleshoot or recover a session recorded that way (Rooms overview). It runs in Chrome and other Chromium browsers, Firefox, and Safari 26 or later (Rooms FAQ).
• The hard words. Product names, guest surnames, the acronyms your sector runs on. Step 1 needs them.

1. Load the glossary before anything is transcribed

Open a project, click the menu in the top-left corner, choose Tools, then Transcription Glossary, and type each term on its own line, as the glossary article shows. It holds 30 terms per drive, covers English only, and adds a word by itself once you have corrected it three times. Every project on the drive shares it, so one list serves the whole show.

2. Record the episode in a Room

From the Drive view, click Record, then Record with others, following the host guide. Click Invite and copy a link for each role: Guest is recorded with no controls, Co-host is recorded with shared control, Producer has full control and is not recorded. Approve each guest as they ask to join, or turn on auto-approval in Room settings.

When the session ends, each participant's high-quality file uploads from their own machine, and Descript builds a sequence, places it in a composition and transcribes it, per the Rooms editing article. Browser backups arrive first, and you can start on them. Recorded on Zoom instead? Connect Zoom under Settings, Connected apps, and import the cloud recording; Descript recommends a separate audio track per participant (Zoom import).

3. Name the speakers

Room recordings take their speaker labels from the names people used in the Room. For an imported file, Descript detects distinct voices during transcription, for files up to 5 hours. Click Identify speakers, play each sample and type the name, per the speaker detection article.

4. Cut the conversation like a document

Select a passage and press Delete on a Mac or Backspace on Windows, and the audio under it goes too. When you are unsure, press Command + Delete or Control + Backspace instead: the text is struck through, stays visible and is skipped on playback, per Deleting vs ignoring. Both can be undone. Choose with the export in mind, because ignored text can still turn up in an exported transcript while deleted text cannot.

5. Take out the ums and the dead air

Open AI Tools in the right sidebar and choose Remove filler words. For each one you can delete it, delete it and leave a gap clip, ignore it, or remove it from the transcript only; Avoid harsh cuts skips any that cannot come out cleanly (filler words article). Then run Shorten word gaps, set what counts as a gap and the length to cut it to, step through the results and click Shorten or Shorten all (word gaps article). Both draw AI credits.

6. Clean the sound

Select the script layer, open Properties and toggle Studio Sound under Audio Effects. For a Room recording with a track per guest, open the sequence in the Sequence Editor and enable it track by track (Studio Sound article). It applies to the file everywhere it is used, and the Intensity slider sets how hard it works.

7. Mark the chapters

Type # in the script, click Add marker on a blank line, or run Add Chapters from AI Tools (markers article). Descript web links show them as chapters by themselves; for a local file, turn on Include markers as chapters under Metadata. YouTube ignores chapter data inside an uploaded file, so use Copy chapters for YouTube and paste the list into the description.

8. Let an assistant draft the notes, and listen yourself

There are two documented routes. Inside the editor, the Publish tools draft a title, a summary, show notes with chapters and timestamps, and social posts from the finished script (Publish tools). Or add Descript to Claude from the connector directory and work from a chat; the Claude setup article gives prompts such as "Add Studio Sound and remove filler words from the second composition", and the MCP exports transcripts as Markdown for the assistant to turn into notes (MCP overview).

The human check is the step that cannot be skipped: the assistant drafts, and a person plays the edited episode once at normal speed and reads the notes against it before anything is published. The Underlord article says the co-editor may overpromise or make wrong assumptions. The MCP is also scoped to one drive and cannot hand over an MP4 without publishing to a web link first.

9. Export the episode

Choose Local export, then MP3, WAV or M4A. Under Advanced you set channels, sample rate, bitrate and volume normalisation, with targets such as -16 LUFS; under Metadata you add show title, episode title, artwork and chapters, which WAV files cannot carry (audio export article). If your editor works in Premiere Pro or Pro Tools, timeline export keeps the edits, on Creator, Business and Enterprise (timeline exports).

What the help centre flags

• A guest on a phone. Unsupported in Rooms, and unrecoverable if it fails.
• Studio Sound on a very loud background. The waveform can go flat; lower Intensity or switch it off for that file.
• Filler words in other languages. Detection covers English transcripts only.
• A stalled upload. The guest rejoins the Room to finish it, or uses a recovery link, and you keep editing on the backup meanwhile.

How to know it worked

Take a baseline from the last episode edited the old way: hours from recording to published, and whose hours they were. Log the same for the next two episodes in Descript. Then open the Usage tab, which exports a CSV on paid plans, and divide the AI credits used by the episodes edited (usage article). That is your credit cost per episode.

Set that budget before you lean on Underlord. The usage article says each prompt is charged for its reasoning and again for any tool it runs, and that two similar requests can cost different amounts. A co-editor billed per request has no reason to suggest you stop asking, so the stopping rule has to be yours.

FAQ

Does a Room with five people use five times the minutes?

No. A Rooms session counts its own length: one hour with five participants uses 60 media minutes, not 300, per the usage article.

Can I edit the Room recording before every upload has finished?

Yes. Backups appear first and you can work on them; once a guest rejoins or uses a recovery link, the full file can be added, per the Rooms editing article.

Will my chapters show up on YouTube?

Not from the file. YouTube does not read chapter metadata, so copy the formatted list from the marker menu into the video description, as the markers article explains.