Deepfakes are often synonymous with threats. We repeat this regularly at Citrus Juicer because, while this technology is not immune to certain abuses. The most recent example was provided by this joke, in which a misrepresentation by a Russian opponent deceived several MPs. The consequences were limited, but we can envision in the future that these creations could undermine the very foundations of our democracies.
However, deepfakes can also enable advances, especially in marketing. An example of this is provided by the artificial intelligence Marvel.AI from Veritone. This tool, identified by DigitalInformation, is based on a very simple operation.
Audio deepfakes for personalized advertising?
Specifically, it is first necessary to restore real records of a person’s voice. Once that base is reached, the AI can create sounds based on different areas of emotions and intonations. The applications of this technology are immense.
Our colleagues cite the example of celebrities or influencers who do not have the means to record and in this way can understand certain statements without having to utter a single word. This functionality can also be used in conjunction with the many audio formats available online: podcasts, conferences, etc.
The question of possible deviations remains. Veritone thought about securing its tool by making sure the language library was never open to the public. It also adds certain inaudible watermarks to the sounds as well as the ability to protect and claim the use of the voice to avoid theft or theft.
This company is by no means the only one working on audio deepfake technologies. For example, last year we told you about California-based startup LOVO. This company has developed a tool with which you can precisely and convincingly recreate vocals from a short audio excerpt. Managers already envision that in the future it will be possible to create targeted ads that “call every user by their name”.