Monday, April 24, 2023

AI Voices & the Current AI Controversy

Little did I know, when I started to create the new Moment of Inspiration training series that the decision to use AI voices would change my release plans so dramatically.

From the beginning, I had planned to upload the Moment of Inspiration Course onto Udemy.  But, little did I know that including AI voices to supplement my own narration would be cause for rejection by the Udemy team.

The reasons for Turning to AI Voices

Ten years earlier, when I created my first Moment of Inspiration Curriclum, my voice was strong enough to complete hours of video.  By the time I started on the new curriculum I was ten years older and, as it turned out, was having age related heart issues.  My voice simply no longer had the stamina to smoothly narrate all the videos required to teach Moment of Inspiration effectively.  

I did not want to give up on the project because I know, without a shadow of a doubt, that at-risk young students' lives have been radically altered through the 3D design and printing classes we taught at YouthQuest.  So, I wanted to continue that work if at all possible.

The first attempts at finding suitable voices to use didn't go all that well.  While Descript allowed me to use my own voice, I never really seemed to evoke the right tone behind the words.  Speechelo was OK.   But, again, I was not able to control the voices to my liking.

Fortunately, I found WellSaid Labs.

The voices I ended up using not only conveyed the pronunciation; but, the way WellSaid Labs Studio worked, allowed me to tailor both the pronunciation and the inflection to a point where the narration is 95% or above exactly as I would have spoken it.

I am not trying to fool anyone by using voice other than my own.  I simply needed a script to be narrated at the lowest cost possible while still being effective.

The Controversy

 The rejection of a course simply because it includes AI voices, has to be seen in the broadest context.  We have jumped into AI so quickly that the full ramifications of the technology are totally a mystery and the potential for harm, in the broadest sense, are certainly there.  I assume that Udemy is simply trying to get ahead of the Intellectual Property curve and it's easiest to simply develop a blanket ban while the technology is being sorted out.

But, from this one small corner, I would like to address my own experience with using WellSaid Studio as part of the workflow of my course development.

Using AI Voices Improved the Content

I have been creating video content since 1969, when I begin with a used Sony TVC-2010.

One of the realities that has remained the same in all of those years was how difficult it has been to match sound levels between takes.  It was a VERY pleasant surprise to discover just how seamlessly it sounds when new AI material is inserted into previously recorded segments.  This allows for improvements and additions to scripts in a way unmatched by live recording.  I found my narrations getting better in the editing process because I could reflect on what had been there and make it better and more clear so easily.

The other aspect of using AI voices that I believed helped the content is that it put me into the role of director, rather than simply writer/narrator.  Hearing the words I had written in a voice other than my own allowed me to refine what I written and work harder on getting the inflection to convey my thoughts.  It was a remarkable experience.

Using AI Improved the Flow

Designing in 3D is a multi-step process.  Being able to change narrators to indicate a new series of steps or a new focus of the course content was another benefit of using  AI voices to augment my own narrative.  

The two voices I most used were identified as Ava M. and Wade C.  But, even within those voices I was able to select different styles based on the subject matter being delivered.  The only issues I had as I switched from voice to voice is that Wade C., using the "Conversational" style was considerable louder that the other voices.  But, this was easily dealt with using the Limiter and Loudness Normalization effects in Audacity.  


Quite frankly, it was a big bummer for me when Udemy would not allow my course simply based on the fact that I had used AI voices.  But, in the grand scheme of things, that loss pales in comparison to the benefits that AI voices brought to the project as a whole.  I don't know how negatively AI might impact society in the long run; but, I can attest to the terrific value of AI voices in helping me overcome a disability to allow me to complete something very important to me.

The new 3D Design for Fun and Life Course.

No comments:

Post a Comment