How to Add a Speaker in Descript

You hit record. The conversation goes great. Then you open the transcript — and it’s a wall of text with no idea who said what.

That’s the exact problem speaker labels solve in Descript. Whether you’re editing a podcast, a video interview, or a multi-person recording session, knowing how to add and manage speakers inside Descript turns a chaotic transcript into a clean, editable, professional script.

This guide walks you through every method — from auto-detection to manual labeling, AI voice speakers to default settings — so you never lose track of who’s talking again.

What Are Speaker Labels in Descript?

Speaker labels are identifiers attached to sections of your script that mark who is speaking at any point in your recording. They do more than organize text — they are essential infrastructure inside Descript.

Here’s what speaker labels enable:

• Script organization — Every line is attributed to the right person, making multi-person transcripts readable at a glance.
• AI speech generation — Speaker labels are required before you can generate AI audio using Overdub or any AI voice feature.
• Copy-paste continuity — When you copy and paste content across compositions or projects, Descript carries the speaker label with it automatically.
• Color-coded clarity — Each speaker label gets its own color in the script, giving you instant visual context as you edit.

With Descript’s transcription accuracy reaching up to 95% under optimal conditions, the platform already does heavy lifting. But speaker labels are what turn a raw transcript into something you can actually work with.

How Descript Auto-Detects Speakers

The fastest way to get speaker labels into your script is to let Descript do it.

When you upload an audio or video file and transcribe it, Descript’s AI automatically analyzes the recording and assigns speaker labels to each detected voice. It identifies how many unique voices are present — whether that’s 2, 5, or 12 speakers — and segments the transcript accordingly.

Once detection is complete, your job is simply to name who’s who. Descript’s Speaker Detective feature walks you through each identified voice, plays a sample, and lets you type in the name. You can use Enter to move between speakers without touching your mouse, keeping the workflow fast.

This AI-powered detection was introduced to address one of the biggest friction points in podcast production: manual speaker labeling is time-consuming, and automatic transcription alone doesn’t solve it. Descript became one of the first transcription services to offer multitrack transcription with automatic speaker labeling — a milestone that significantly changed multi-speaker editing workflows.

To get the most accurate auto-detection:

• Position yourself 6–10 inches from your microphone during recording
• Speak clearly and minimize background noise
• Use separate audio tracks for each speaker when recording remote interviews

How to Add a Speaker Manually

Auto-detection handles most situations, but sometimes you need to insert a speaker label by hand — for example, when adding a new section to an existing script, or when auto-detection misses a transition.

Here are your options:

Option 1 — Use the @ shortcut

Click inside the script where you want to insert a speaker label. Type @ on your keyboard. A dropdown will appear. Either select an existing speaker from the list or type a new name and select Create speaker ‘name’.

Option 2 — Use the Actions menu

At the top of the script, click Actions…, type “speaker” in the search box, and choose Insert speaker label. From there you can select an existing label or create a new one.

Option 3 — Use the three-dot menu

Click the three horizontal dots (…) icon at the beginning of a blank line and select Add speaker label.

All three methods accomplish the same thing. Use whichever fits your keyboard or mouse workflow.

How to Rename a Speaker

If Descript auto-labels your speakers as “Speaker 1” and “Speaker 2,” renaming them is the first editing step.

Click any instance of the speaker label in the script. Start typing the correct name directly. Descript will update that label in that section.

If you want to rename every instance of a speaker across the entire project at once, use the Replace in project with… option:

1. Click any instance of the speaker label
2. Select the speaker options button (the three-dot icon to the right of the label)
3. Choose Replace in project with…
4. Select the correct speaker label to replace it across the whole project

This is the fastest way to clean up auto-detected speaker names at scale.

How to Reposition a Speaker Label

Speaker labels don’t always land exactly where they should. If a label is off by a word or two, you can drag it to the right location.

Hover your mouse over the speaker label in the script. A reposition icon will appear to the left of the name. Click and drag the label to its correct position. As you drag, the words in the script will highlight in blue — these are the valid drop zones. Release the label on the word where the speaker begins talking, and the label will move to that position on a new paragraph line.

This makes correction fast and precise, especially for recordings with rapid speaker transitions.

How to Change a Speaker’s Label Color

Every speaker gets an automatic color code for visibility. If you want to change it:

1. Click any instance of the speaker label in the script
2. Click the speaker options button (three-dot icon)
3. Select Change label color
4. Choose a new color from the list

Color-coding is particularly useful in dense multi-speaker transcripts where visual scanning saves time.

How to Remove a Speaker Label

Need to clean up a speaker that no longer applies to a section?

1. Click any instance of the speaker label
2. Open the speaker options button
3. Select Remove from project

Once removed, Descript reassigns that speaker’s sections to the speaker appearing directly above them in the script. This keeps the transcript intact without leaving any orphaned text.

How to Create an AI Speaker in Descript

This is where speaker labels move beyond organization and into production power. Descript allows you to create an AI clone of any speaker’s voice — which can then be used to generate new speech directly from text. This is the Overdub feature.

Here’s how to set it up:

1. Add a speaker label to your composition — click Add speaker at the top of your composition
2. Select Create speaker from the dropdown and name the speaker
3. Click the new speaker label, find the name in the dropdown, hover over it, and click the … menu
4. Choose Enable speech generation — this opens the consent and authorization flow
5. Select your microphone and click Record
6. Read the consent statement clearly and naturally (must be in English, even for non-English projects)
7. Stop the recording, review, re-record if needed, and click Submit

Your AI Speaker will be ready within minutes. Once created, you can type corrections directly into the script and Descript will generate the audio in that speaker’s voice — no re-recording needed.

You can also create AI Speakers from the AI Speakers tab inside your Drive view.

How to Add a Stock AI Speaker

Don’t want to record your own voice? Descript offers a library of pre-built AI voices you can assign to any section of your script.

1. Click the Add speaker label option at the top of a paragraph (or use the @ shortcut)
2. Select Browse stock AI speakers from the dropdown
3. Preview voices by clicking the play button next to each one
4. Select the speaker that fits your content — each is tagged with descriptors like “Soothing and narrative” or based on language support

Important: For the best results, use a stock speaker whose tagged language matches the language of your script. Descript supports AI stock speakers in a wide variety of languages. Audio may still generate if languages don’t match, but the result may sound less natural or lose the intended tone. To see non-English speakers, click the filter icon in the speaker list and select Show non-English speakers.

How to Set a Default Speaker for New Projects

If you consistently use the same AI speaker across projects — especially when starting in Writing mode — setting a default speaker saves you the step of assigning one every time.

1. Open Descript and go to app settings
2. Navigate to the General tab
3. Find Default Speaker
4. Select your preferred speaker

This setting applies when beginning a blank project, not when uploading existing media. Once selected, Descript saves the preference automatically.

Why Speaker Labels Matter More Than You Think

Most people add speaker labels just to organize their transcript. That’s only part of the value.

With over 546 million monthly podcast listeners globally as of 2024 — and that number projected to keep climbing — audio and video content production is accelerating fast. More recordings means more transcripts, more editing, more complexity. Speaker labels are the infrastructure that makes it all manageable.

Here’s what properly labeled speakers unlock beyond basic organization:

• Faster editing — Jump directly to a specific speaker’s sections without scrubbing through the full timeline
• Cleaner outputs — Export transcripts with proper attribution for show notes, interview summaries, or meeting documentation
• Voice isolation — Apply audio cleanup, formatting, or effects to individual speakers without touching others
• Separate clip generation — Pull clips featuring only one participant without manually hunting through timestamps
• AI speech generation — Speaker labels are required to enable Overdub and text-to-speech workflows

Descript’s transcription is up to 95% accurate in ideal recording conditions, but it handles multiple speakers, diverse accents, and technical vocabulary consistently well. The text-based editing approach — where changes to the transcript directly modify the underlying audio and video — means that speaker-labeled transcripts become the primary editing interface, not just a reference document.

The industry average transcription accuracy sits between 85–95%. Combined with Descript’s speaker detection, you’re getting a near-complete first draft of a multi-speaker recording without any manual work at the start.

Common Speaker Label Issues (and How to Fix Them)

The wrong speaker is assigned to a section. Drag the speaker label to the correct boundary using the reposition icon. The script will update instantly.

Speaker labels are mixing up voices. This happens most often when two voices sound similar or when there’s significant background noise. Manually reassign the affected sections using the @ shortcut and correct speaker name.

I need the same speaker name to apply everywhere, not just one section. Use Replace in project with… from the speaker options menu. This updates every instance in the project at once.

My AI Speaker isn’t ready yet. AI Speaker creation typically takes a few minutes after submission. If it’s been longer, check that the consent recording was submitted successfully and the audio was clear.

I want to remove a speaker but keep their text. Use Remove from project from the speaker options. The text will automatically be reassigned to the speaker above that section in the script — nothing is deleted.

Conclusion

Adding a speaker in Descript is one of those features that looks simple on the surface but runs deep once you understand what it enables. From auto-detection that identifies every voice in your recording to AI speech generation that lets you fix lines without touching a microphone, speaker labels are the foundation of Descript’s entire editing workflow.

Start with automatic detection for any multi-person recording. Use the @ shortcut to add or correct labels manually. Rename speakers immediately after transcription so your script is readable from the first edit. And if you’re building content at scale, set up AI speakers to handle corrections without re-recording.

Once your speaker system is clean, everything else in Descript — editing, repurposing, exporting — becomes dramatically faster.