Will AI voices replace humans?
Updated: May 3
I have followed the progress of AI voices for more than a decade now. 10 years ago the former Canadian company Lyrebird managed to make AI voices sound quite realistic. Their technology also introduced the possibility of “sampling” the timbre/character of a person's voice and placing it on another voice recording. You have probably heard this technology in use in "deep fake" videos with Obama and Trump.
Back then, I thought AI voices would take over in just a few years and that human voices would only be used in prestige projects. I was wrong. We are still suffering from audio artifacts and mechanic reading in AI voice services.
However, recent AI progress has made results so good that I’m starting to wonder.
- If the result from a machine is believable, does it then matter if it’s conscious or not?
- If AI can produce results that we see as creative from simple text input, is it then creative?
The answer is not obvious and I guess in many cases the results will be useful and just as good as if they were made by a human.
But, even if we now see many convincing demos from AI-driven services, most of them still have had some help from a human hand to be that good. For example, an audiobook read by DeepZen AI can sound quite good but will take weeks to produce as it has to be manually checked and edited by a human.
Instead of being replaced, we will probably work tight together with technology and many times very different from how we work today. New technology will create new possibilities and new job opportunities. It's not the first time in history that automation has taken place and we still have to work.
To sum up. AI voices still, even though in a very advanced way, just analyze words and phrases and therefore make mistakes in mood and emphasis. This makes it far more time-consuming to work with AI voices than humans in most cases. At least if you want a result comparable to humans. However, they are already useful for many less expressive readings, interactive real-time services, temp voices, etc. We will also see many new interesting possibilities when the algorithms get even better.
The future is interesting and the only thing we know for sure is that we don’t know what it will look like. And if AI gets conscious for real, then I guess we all have a problem ;-)
You can try some AI voices in Voicemachine for free. Just start a project, copy and paste your script, choose AI voice and listen to or download an audio file. Works for scripts up to about 500 words (max 5000 signs).