When test subjects in a new American study were asked to listen alternately to human and AI-generated voices, without knowing which was which, many listeners preferred the AI version, reports Variety.
Storytel already offers its users a feature that allows them to switch between listening to a human narration and an AI-generated voice for certain titles. However, the audiobook giant's position is crystal clear: AI should not replace human narrators.
And what we see is that in cases where we have a "voice switcher", around 90 percent choose the human voice, says Helena Gustafsson.
Short readings only
The content manager takes the American study - commissioned by an AI audiobook company - with a pinch of salt, especially since the test subjects only had to listen to short readings.
All studies comparing or looking at short audio clips with AI voices get high ratings. It is when using AI voices for entire audiobooks that listeners tend to react because an AI voice cannot reproduce the same dynamics and tonality.
This does not mean that Storytel is against AI readings. On the contrary, Helena Gustafsson believes that the publishing industry will benefit from the cheaper AI method, enabling even books that are not expected to attract a large audience to become audiobooks.
In the Swedish market, this applies primarily to non-fiction books. If you can make a segment that is not currently popular popular through AI reading, you can later read the books that have become successful with a human voice.
Clone actors
Another way to go is to AI-clone the voices of popular narrators, which Storytel has done with the voices of actor Stefan Sauk and Sune author Anders Jacobsson. The latter "out of necessity", as after suffering a stroke he was unable to read the young adult novels that are strongly associated with his voice.
Looking ahead five years, how many audiobooks do you think will be read by AI?
As a speaker, I wouldn't feel threatened, but AI voices are here to stay and will become more numerous. What determines to what extent is not the technology itself but how the discussion in society looks and how legislation develops.
