According to a recent announcement from Amazon, the Alexa digital assistant can now theoretically mimic any person's voice using a few clips, or around one minute of recording.
If it performs as promised, it would be a technological breakthrough because computer-generated voices still don't pass muster as human voices and aren't the best option, for example, for video voice-over. The live demo may be seen in the following video at timecode 1:02:38:
In practice, I discovered that it simply changed the tone and voice print of the normal machine-generated agent to match the desired human voice print. Without knowing the original voice, it is difficult to assess how well the example worked, although it seemed to work very well while still being somewhat artificial.
The chosen line was unquestionably a good choice for the demo because it lends itself to a sluggish, somewhat robotic reading. When applied to an audio stream, the technique is comparable to the AIs that Picasso utilized to convert your photos into paintings.
Having Alexa talk in your favorite friend's, family member's, or celebrity's voice may sound like fun. However, utilizing audio recordings from distant family members has swiftly gained popularity on the Internet. That was the use case presented in the video above by an Amazon official.
On the one hand, hearing the voice of a loved one who has passed away can sound like a therapeutic experience. It might be a slippery slope with unanticipated effects, though. Many people began to wonder whether voices might be utilized without permission and whether technology could be used to mimic real persons.
Depending on the circumstance and the users, the response is usually "it depends." One thing is certain, though: these technologies already exist and will continue to advance. Synthesized voices will soon be impossible to tell apart from human ones.
Post a Comment