Art, Text — What’s Next?
Now we all know there is something simply irresistible about the advent of AI in our daily lives. It spreads like a very-well-known virus infecting so many articles, posts, video platforms in addition to every conversation imaginable between a pair of people.
With so many raging visual artists and writers concerned about the ethics and consequences of machine-generated content, the other question of whether or not AI could ever be sentient is somewhat being thrown under the bus.
Another advancement underway happens to be in the voice-over industry – so let’s start by assessing the current state of AI voices.
Nowadays, it boils down to synthetic, robot-like voices with limited inflections, tonality and fluidity. The process of imitating human voices is much like text on a computer versus handwriting (which is riddled with so many imperfections and inconsistencies — similar to our natural way of speaking). This is most definitely not the case with typed text: making it somewhat soulless and unoriginal.
When we distinguish between static and dynamic content, synthetic voices already have a role in navigation devices and voice assistants. It appears that no consumer minds too much, at least according to a study by Capgemini which states that 74% of consumers utilize conversational assistants in one way or another for purchases. When it comes to navigation services, however, I think we are more concerned with the ridiculous, time-consuming paths the apps force us to take.
Get Jana Osman’s stories in your inbox
Join Medium for free to get updates from this writer.
With static content such as video or audio intended to inform or entertain, people are generally far less accepting. Despite there being a TikTok trend sort of normalizing robotic voice overs, it lacks that distinct ‘spark’ that makes “entertaining content” entertaining. It is also true that the popularity of unnatural voices has been around for a while, with software such as Vocaloids — which are user-generated songs created with Japanese software that generates artificial “singing voices”. Albeit the lack of human emotional vocal components, they manage to receive interesting responses such as:
as well as:
…in addition to many other commenters who appear to share the same sentiment.
Its viability is very likely to increase in the future as standards improve, though.
Machine learning works by setting up parameters on different features from what it intends to study to ‘train’ it – this information then being stored and used to regenerate similar pieces that incorporate those learned features. AI can then grow and adapt using its current ‘library’ of knowledge and expand using further sources.
Nevertheless, working with a professional voice acting organization such as Voice Talent Online can benefit many companies by providing reinforcements such as quality mixing, editing, and consultations for better solutions tailored to the company’s needs and marketing strategy. After all, voice talent is much more than simply lending a voice – an actor sells their image, brand and personality which could really make it or break it when trying to gather an audience for an animated piece or commercial.
With so many deepfake technologies roaming around, it might just be that actors will have to trademark their voices fairly soon. Who knows? Maybe other unprecedented hallmarks of human creativity – or even features of our bodies – will have to be trademarked next!
This article was written for Voice Talent Online’s blog contest — Scholarship #2
