Video software provider Vyond has partnered AI voice-over company Wellsaid on improved genAI for audio content creation.
Consequently, a new blog has Vyond customer success manager Craig Kingsbury explaining how to make the best of today’s AI voice generation technology.
“Many of today’s are modelled on real voice actors via ethical partnerships. So they’re smooth and natural-sounding. And often impossible to tell from human voices,” he said.
Because AI voice generators can sound robotic and awkward, they’re not always considered good replacements for human voices.
But AI voice generation can save money, Kingsbury said. When the quality is high, the right content can improve audience engagement, helping change behaviours.
As a result, Vyond now offers multiple AI voice options for customers to use when making an audio track for their videos. That includes AI voices from WellSaid, integrated with Vyond’s Enterprise and Agency subscriptions.
Vyond and Wellsaid tips for content creation
Kingsbury and Wellsaid’s Nate Grieb present how to get more from AI voice-overs in this webinar. For example, translation into different languages is easier. Branding is consistent. And it can be quicker, because there’s no studio time to book or pay for.
“Whether for training, marketing, communications, or sales, quality and authenticity of voices (can) make all the difference.”
However, sometimes results can be disappointing. A first take might have a flatter tone and delivery. In some cases, the AI voices might use odd or incorrect pronunciation. Or the speed is slower or faster than you’d like.
Yet vocal inflections such as speed, pitch, and pauses can be adjusted, and pronounciation fine-tuned, Kingsbury wrote.
Also, when writing your script, think about how people naturally speak.
Typically, short sentences make for cleaner, smoother narration that’s easier to understand. And don’t forget to use punctuation to create emphasis, pacing, and pauses, he said.
Choose a suitable AI voice for your specific goal. Consider tone and style. For example, do you want a conversational sounding voice, or something more authoritative?
To change pronunciation, try using quotation marks around the syllables you want stressed or emphasised.
Or try writing words as they sound; for example, writing Saskatchewan as ‘sask-ACH-oo-when’. And Wellsaid has a guide with further suggestions.
Spell things out differently for better results
In addition, acronyms and numbers can be hard for AI voices. Again, try spelling out acronyms phonetically. L&D might work as ‘Ellen Dee’, for instance.
“AI voices are trained on broad vocabularies, but they may misinterpret words,” Kingsbury wrote.
“You can also try using dashes, full stops, or spaces for letter-by-letter pronunciation, like for ‘T-B-D’ and ‘A.S.A.P’. If by ‘2025’ you mean the year, try spelling it as you’d like it pronounced: ‘two thousand twenty-five’.”
Finally, working with shorter audio clips can make editing easier to home in on issues. Hone your content as needed, Kingsbury urges.
Vyond explains more about how and why to make a video tutorial here.
( Photo by Gregory Cole on Unsplash )
