In the age of fake news, deep fakes are a new kind of media manipulation that’s become increasingly popular.
But what if you could use it to your advantage? This is what Descript is aiming to do with the feature of their Pro plan called Overdub.
Check out what this article sounds like using my overdub voice and let me know what you think
What is Overdub?
Overdub creates a computer model of your voice from recordings that you submit. This allows you to replace words in your transcript and have those changes reflected in your audio recording. Overdub also can create completely standalone audio. Which means that you can use it both for editorial corrections and long form narration. My experience with it is that it sounds better in short snippets than long paragraphs that give the listener more time to realize that something is off.
How does it work?
The way that Overdub works is that it has you read between 10-90 minutes of a script that they provide. For me that script seemed to be the voice over for a nature documentary, but your mileage may vary. It is important that you read it with the same kind of energy that you want the Overdub voice to have.
By submitting the reading of a known text to them they can break those words into the different phonemes. Phonemes are the individually distinct sounds that distinguish one words from each other. In the English language there are 44 different phonemes, but there are also tones and inflections which overdub capture as well. It takes about 12-24 hours for your Overdub voice to be finished.
Where should you use it?
This technology is amazing and if you are voice over talent within a few years this will be something to worry about. However, the best use case that I can see for this technology as it is right now in Descript is for fixing small mistakes in something a podcaster said. Did you say that your book sales were going live on Tuesday, but it isn’t happening till Wednesday? Delete the word Tuesday, type Wednesday, and let Descript fix that for you. Did your Zoom recorded call (that you are totally not doing because you know better) get garbled? No problem, you know what you said, type in the words and let Descript fix that for you.
I’m excited about this technology and it has the potential to be a game changer in the podcasting space but just like a photoshopped image you can notice some inconsistencies when you look or listen closely. If Descript Pro for transcriptions and audiograms is already on your radar, then I recommend adding Overdub as well. It gives voice recordings more flexibility by allowing the editor to fix what was originally said without getting the voice talent back on the microphone.