Today, I will walk through one of Descript’s coolest features for content creators: Overdub.
To get started with Descript’s Overdub feature, you have to submit your voice sample for training, and then the tool creates a text-to-speech model of your voice. From there, you can type in sentences, and a voice-over will be generated automatically of your very own voice.
Note: If you’re not looking to clone yourself, you can also use their library of high-quality stock voices too.
So how exactly does this feature work, and what are its pros and cons? You’ll find all your answers in this in-depth Descript Overdub review.
What is Overdub In Descript?
Overdub is a special feature of the Descript video editing software that lets you clone your voice to create custom voice-overs by typing text in the editor.
It uses Generative Adversarial Networks (GANs) to match your voice tone and synthesize it with general human speech patterns to generate high-quality voice output.
Descript also allows you to use its dozens of high-quality stock voices for generating and editing audio content.
Here are some possible use cases for Descript in content creation:
- Voice actors can use it to read out repetitive portions of their audio project.
- Writers can become podcasters and can use it to create impressive-sounding long audio recordings without the expensive setups
The biggest benefit of Overdub is that you can correct voice recording mistakes by simply editing text. Before this, you were forced to record the whole audio again or trouble your editor to cover up the mistake.
How do you make an Overdub in Descript?
I’ll walk through the entire process of creating your custom Ai Overdub voice and using it for a given voice sample step-by-step.
A quick disclaimer: Overdub is available only in English for the time being, and you can clone only your voice.
What Do I Need To Create Overdub Voice?
You need to submit at least 10 minutes of sample speech to the Descript Overdub portal so that it can process your voice within 1 day and generate your custom overdub voice. (It is recommended to submit 25 minutes + if possible)
Follow these tips to record a better-quality voice for sample speech:
- Record your voice using a high-quality mic in a quiet room without background noise. Or, more importantly, record in the studio where you plan to record your final content.
- Record between 20-50 minutes of your voice. Submitting hours of sample speech will worsen the voice output.
- Speak as you would normally record for your audio and video content. The more natural your sample, the more natural your overdub will be.
- You must submit your voice ID and formal consent to use your voice for the overdub. Ensure your voice ID matches your sample speech.
Here’s the best part:
You can upload multiple audio files for your sample speech as long as they sound similar and are registered in the same setting. In addition, descript overdub is smart enough to handle occasional mispronunciations in the sample, so you can continue recording despite speaking errors.
How To Submit Training Audio To Overdub?
Download the Descript AI software on your PC and follow the steps below to submit training audio to generate Descript studio sound:
- Sign in using your account to find the drive view.
- Select “Voices” from the left side of the Descript Drive View.
- Click on “Create a new voice.”
- Enter a voice name and click on “Confirm.” You’ll be redirected to the Overdub voice window like this ????????
- Next, click on the mic icon at the top and then the red circle icon to record 30-40 minutes of voice recording following the tips I mentioned earlier.
- You can also upload an existing audio file from your computer by clicking on “Add new” > “File from computer” in the left-hand corner of the overdub voice screen.
- After you add your training audio, click “Submit training data” in the upper right-hand corner of the screen.
- You have to record and submit a Voice ID to give your consent to use your voice for the overdub feature. Your voice ID should match the training data audio, or else the overdub request will be rejected.
- Descript will send an email within 2 to 24 hours once the overdubbed voice is ready.
How To Use Overdub Voice In Your Script/Composition?
Follow the steps to use your custom overdub voice in your composition:
- Open Descript Drive View > Recent Projects.
- Click on Speaker Label and select your overdub voice.
- Once your overdub voice is assigned to the speaker label, you can start writing your script in the descript editor window. You can also upload your script from the computer.
- Wait a few moments for Descript AI software to generate your Overdub voice.
You can also make corrections using Overdub voice by highlighting the wrong portion > selecting overdub from the hover menu (as shown in the screenshot) > typing in the correction > and selecting overdub again to generate custom voice correction.
How Do You Use Overdub Stock Voices?
Using overdub stock voices is similar to using your own voice for a given word document.
Follow the steps below:
- Head to Descript Drive View and click “Recent Projects” in the left-hand corner. Select any of the projects you’ve saved with your script.
- Click on Speaker Label > Stock Voices and choose from 9 available options (as shown in the screenshot). You can click on the play icon to check how each stock voice sounds.
- Once you’ve chosen the stock voice, hit Enter and wait a few seconds to generate the overdub stock voice recording!
Is Descript Overdub Good: Pros and Cons
All and all, Descript Overdub is an amazing audio editing tool. You can shorten your learning curve in editing audio and video to a great extent and create amazing voiceovers without professional training.
It is nothing short of magic when it comes to inserting quick punch-ins of forgotten words. However, longer sentences could still use a bit of humanization and improvement. That said, the program is bound to only get better, and it’s still pretty cool to make and hear your own Ai voice clone.
I also found its stock voice options way better than Speechify, Google text-to-speech, and Speechelo. You can easily use Descript’s 7-day pro trial to see if it works for you.
Descript Overdub Review: FAQs
Overdubbing, in the present context, refers to voice cloning technology that allows you to create a model of your voice to generate voice-overs on any given text. Research in text-to-speech started in the 1980s, with the earliest version of voice cloning applications being Apple’s Siri.
Today you can submit your voice sample to dub your voice for hours of recording scripts using ai tools like Descript, Amazon Polly, and Real-Time Voice Cloning. We recommend Descript for better audio output and easy-to-use UI.
Descript Overdub uses Lyrebird AI based on the Generative Adversarial Network to generate natural-sounding audio from a given voice sample. It utilizes deep machine-learning frameworks to understand the general pattern of human speech from thousands of hours of training audio.
Once you submit your 30-40 minutes of voice sample, it matches its general understanding to find uniqueness in your voice to clone your tone for a given script.
Descript’s free version allows 1000-word vocabulary overdubbing with “um” and “uh” filler word removal and 10 minutes of studio sound.
Their pro plan starts from $24/mo, with unlimited overdub, over 18 filler, and repeated word removal. In addition, you can contact the Descript team to create a custom Enterprise plan for studio house requirements.
Descript offers 9 different natural-sounding stock voices in its Overdub feature, with 5 male voices: Don, Malcolm, Ethan, Henry, and Nicholas, along with 4 female voices: Emily, Carla, Ruth, and Nancy. You can play each voice before using it in your script.
Different descript overdub alternatives include Amazon Polly, Google Speech Services, SpeechParrot, Real-Time Voice Cloning, and Natural Reader. All of these are great if trying to start a podcast. I found Descript overdub custom and stock voices much more natural than competitors, but you can always use their 7-day pro trial to find the difference yourself.