Project VoCo by Adobe
With just 20 minutes of prep work, you can have anyone say anything. -Geektime on Project VoCo
At some point of our lives, we’ve all witnessed the borderline sorcery power of Adobe Photoshop. Recently, an announcement at MAX 2016 brought the jaw-dropping power of Project VoCo, known as the “photoshop for audio contents”, into attention. Here’s a sneak peek of its magical power:
As part of Adobe’s Creative Cloud, Project VoCo features state-of-the-art audio editing capabilities. According to Zeyu, who unveiled the product, provided with a 20-minute speech of a person, Project VoCo will be able to generate any word, phrases, and sentences in the his/her voice. Certain concerns are addressed, too. As much effort as Adobe is putting into generating audio that can pass reality check, Adobe is also building “watermarks” in synthesized audio that can be detected when necessary.
The algorithm was not discussed during the reveal, as one would expect. But Project VoCo essentially breaks down a speech into phonemes, and in piecing the phonemes together, “predicts” the unsaid speech. It is really fascinating to me because it is an advancement in fundamental technology, on which so many achivements can build.
Read more about the implications of Project VoCo here.