Overdub overview

Overdub lets you create and edit content using a text-to-speech AI model of your voice or one from our ultra-realistic stock voices. This makes correcting your recordings or generating speech from scratch as simple as typing. This article will introduce you to all the things you can do with Overdub in Descript.

Creating your AI voice

You can use your very own voice in Descript as an AI voice to generate audio or edit your recordings just by typing. To do this, you'll first need to submit a sample of your voice for training and verification.

1. What you'll need

1. At least 10 minutes of recorded speech (audio or video); we recommend 30 to 180 minutes for high-quality results. If you don't have a sample recording, don't worry – you can use Descript to record one. 

2. Permission of the person whose voice is being submitted for training (i.e., you). You will record the consent statement (Voice ID) as the final step when you submit the voice for training.

2. Create a training session

The training session is where you'll add your sample recording the Voice ID to create your AI voice. The training session also has instructions and tips in the sidebar to guide you along the way.

1. Open Descript.
2. Select the Voices tab on the left side of the Drive View.
3. From the Overdub voices window, select + Create new voice.
4. Name the voice and click Confirm.

3. Add your sample recording

Add at least 10 minutes of sample audio or video of your voice, and we recommend 30 to 180 minutes for best results. This needs to be 10 minutes of you speaking. If you upload a 30 minute recording, and you're only speaking for 5 seconds of it, your AI voice will not be created.

There are two main ways to add your sample recording to a training session:

• Import an existing recording 
  • Copy + paste script text from one or more existing Descript projects
  • Import + transcribe one or more existing audio or video files from your computer
• Record directly into your training session; we have a sample script you can read from

4. Submit training session

Once you've added your sample recording, click Submit training data in the top right corner of the editor. Then record your Voice ID — press record and read the consent statement.

Optional: create a voice style

You can adjust the intonation of an Overdub voice by creating various voice styles. Styles typically work best when using Overdub to create content from scratch, rather than replacing existing audio with Overdub.

1. Highlight a section of your script from your training session to use as the source material for your style. Make sure that the total length of the corresponding script is between 5-25 seconds.
2. Right-click on the selected text.
3. Choose Overdub > Create new voice style.
4. Name the style and select Create.

The style will now be available when you're using the Overdub voice in a project.

Write Overdub audio from scratch (text-to-speech)

Before getting started, make sure to create a project or open an existing project.

1. Enter Write mode

Write your script from scratch using Overdub audio in Write mode, which is one of three script modes in Descript. To enable Write mode Press the W key, choose write mode in the top left corner of the script, or simply choose Start writing in a blank composition. 

You'll know Write mode is enabled when the  Write icon appears in the top corner, and there is a blue border around your script editor.

2. Apply speaker label

Speaker labels are used to distinguish between speakers in your script, and to generate Overdub audio.

1. Type the shortcut @ into your script, or click the + button on a script line and choose Speaker label.
2. Choose one of the Overdub voices from the Create from Voice list. You can choose a custom Overdub voice, or scroll down and choose one of our high-quality stock voices.

This will do two things:

• Create a speaker label for your project
• Link an Overdub voice to that speaker label

You can further manage speaker labels in a project by clicking a speaker label, then clicking  Settings in the speaker label dropdown menu.

3. Type or copy + paste in your script

Now start creating! As you type, Descript will begin generating the Overdub audio of the assigned voice. You can even create a full dialogue by using your Overdub voice, a voice shared with you, or one of our stock voices.

If you are copy + pasting in an existing script:

• Some text formatting may not be supported in Descript
• If it is a long script, we recommend copy and pasting the script into smaller sections.

While you type, you'll see blue text with a dotted underline. This means your Overdub audio is processing. Once the underline is gone and an Overdub audio clip appears in the timeline, you'll be able to hear and edit your AI-generated audio during playback.

Editing with Overdub (replace existing audio with Overdub)

Before getting started, make sure to create a project or open an existing project.

1. Get audio or video into Descript

You'll need to add some audio or video into your script and transcribe it. There are two main ways to do this:

• Import and automatically transcribe an existing recording
• Record and transcribe directly in Descript

2. Add a speaker label

Speaker labels are used to distinguish between speakers in your script, and to generate Overdub audio. If you did not add a speaker label when importing a file or recording, you'll need to add a speaker label and assign an Overdub voice to it before you can start using Overdub.

1. Type the shortcut @ on a new line in the script, or click the + button and choose Speaker label.
2. Click Settings in the dropdown
3. Add a speaker label name and assign an Overdub voice to it

This will do two things:

• Create a speaker label for your project
• Link an Overdub voice to that speaker label, so any time you choose to replace existing audio with Overdub, Descript will use that voice to generate a correction

You can further manage speaker labels and linked Overdub voices by clicking the  Settings in the speaker label dropdown menu.

3. Edit your recording

With your speaker label and assigned Overdub voice, you can start using Overdub to make corrections.

1. Highlight a word or phrase you want to replace
2. Press the D key or choose Replace with Overdub
3. Type the word you want to generate, type Return (Mac) or Enter, or press the Overdub button.
4. Your text will turn blue with a dotted underline. This means your Overdub audio is processing. Once the underline is gone, you'll be able to hear your AI-generated audio.

Adjusting Overdub audio

Convert Overdub to audio 

1. Hover over an Overdub clip in the Timeline at the bottom of your editor.
2. Select convert to audio.

Change clip speed

After converting Overdub to audio, you can speed up or slow down the clips from the sidebar.

1. In the timeline, select the clip you want to adjust.
2. In the playback section of the Properties Panel, click and drag the speed value or type in a value.

These adjustments are different from adjusting your composition's playback speed:

• Clip speed changes will be included when you publish or export your composition.
• Playback speed does not affect your published or exported content and only changes the speed during playback within the Descript project.

Add fades

After converting Overdub to audio, You can quickly create audio fades and crossfades by dragging a transition handle in the Timeline.

You can also adjust and apply fades or crossfades by clicking on the transition handle of your clip.