Adobe just revolutionized the sound market by presenting Project VoCo, and it will blow your mind. The new project, already aliased as the “Photoshop for voice”, makes everyone able to edit spoken content in audio recordings, and even adding new content, just amazing!
The app was presented at the Adobe Max 2016 at the beginning of the month, and the Princeton University researcher Zeyu Jin showed how easy it will soon be to manipulate and transform sound files, and even in some cases, to put new words that were never said in the recording. Until now, we could edit audio by manually cutting, copying or splicing together parts of sound waves. Well, with VoCo, this all changes as the program operates in a new principle, which depends on an algorithm that breaks down and recompiles human speech.
The program appears to operate in the principle of identifying phonemes, which are the individual speech sounds we put together to make words and sentences, and logs them, after that, having collected enough data, which Adobe states that takes about 20 minutes of talking, VoCo will have enough phonemes to basically impersonate that person, by joining them to create words and sentences.
In the video bellow, you can see VoCo in action, using a snippet of audio recorded from the comedian Keegan-Michael Key. At first, Jin just rearranges the words, changing “I kissed my dogs and my wife” to “I kissed my wife and my dogs” by copying the words and pasting them in the new order. After that, he adds a new word, changing the phrase to “I kissed Jordan and my dogs”, now that’s amazing. But it doesn’t end here!
After that, he adds more new words, changing the phrase to “I kissed Jordan three times”, now we know it might sound a bit glitchy, but you have to recall, that this is only a prototype version, and therefore, we think that the result is pretty impressive.
Off course that there’s always the issue of the possibility of negatively manipulating digital media, and that can be very dangerous, so Adobe is already working on technologies to embed watermarks and similar in the audio so that in legal cases any tampering attempts will be detected right away, because as we know, these days, the person’s voice is used as a security authentication feature, and this causes a liability.
For now, VoCo has no release date, so we’ll just have to wait, but until then, you can check the awesome video bellow!