Adobe’s new experimental project might change the way we look at, or rather listen to, speeches. Introduced at its annual MAX conference in San Diego, Project VoCo allows you to edit speech like you would a word document. And it isn’t just about editing existing recordings, you can even use the same voice model to create completely new recordings. Project VoCo just needs around 20 minutes of voice samples from a particular speaker. It’ll then analyze and break down this speech into phonemes, transcribe this, and create the voice model. According to TechCrunch, if you listen closely now, you’ll be able to hear when a word is changed but we won’t be surprised if there comes a time when you won’t be able to distinguish the actual recording from the edited/fake one.
Adobe didn’t use traditional speech synthesis technology with this but used what they call “voice conversion.” And it requires hardly any manual intervention. You can edit the auto-generated transcript but there is no need to set timestamps. The algorithms will figure this out themselves. While this could raise a lot of questions, especially since it’ll become harder to trust recordings now. It’s still a pretty cool technological advancement. Adobe won’t commit to shipping this technology but who knows if this’ll show up in any of their products? They’ve done a lot of that in the past.