Loading summary
Transcript1 lines
- [00:00]
A
Foreign I'm David Hinson and I serve as Campus CIO with Bolden Networks for Higher Education. Welcome to Control Alt Lead. This week's episode uses an AI voice clone trained upon hours of my natural speaking voice. While the voice you hear today is cloned, the words, thoughts and ideas here are 100% my own. You may be wondering, why use a voice clone at all? Isn't it easier and faster to simply record your voice directly? Why go to the expense and trouble? Today we'll dig into all of this the problem, the use case, the tools, and the workflow. Then we'll dive into the ethics, policies and practical guardrails that one should consider in deploying AI where authenticity and transparency are fundamental to engagement and learning. This past year, I've been writing, recording and helping produce the podcast that you're listening to now. And while I work closely with our fantastic marketing team on branding, editorial and production, pretty much everything else is me. Even the bumper music. But here's the thing. The podcast is not a formal part of my day job leading Bolden's Executive Advisory Services professionals. It's a labor of love, something I carve out extra time in order to create. Let me explain how the magic is made. My normal podcast workflow starts with collaborating on topic selection, ensuring ideas are timely and useful for our higher ed leaders, directors, VPs, cabinet members and and boards. These often also tie into formal presentations to higher ed industry organizations such as Ace Educause and others. Once a topic is decided upon, the writing begins. Sometimes a draft comes together in a few short hours. Other times it takes days of writing, revising and reworking to get the tone and flow just right before sharing it back with our marketing team for review. That editorial cycle usually takes a week. After approval, I record the episode and the marketing team handles post production by balancing audio levels, ducking in the bumper music, and preparing a file for release. And this is where the biggest challenge always comes in the actual logistics of recording the podcast. A podcast can have the best content in the world, but if the delivery is flat or the audio quality is poor, listeners tune out. At home, I have a professional mic and a quiet studio space, but on the road, that's another story. I have to track down a decent mic, find a quiet room from which to record, and carve out two to four hours to record and edit. What amounts to 10 minutes of polished spoken word recording is the heaviest lift of the entire process. And that, my friends, is where AI voice modeling enters the picture. Several weeks back, I came across a platform 11 labs that could model someone's voice and then generate new audio in that voice with timbre and expression completely intact. Admittedly, I was skeptical. Most tools I'd seen in this space were either robotic, hollow, or sat squarely in the uncanny valley. And not something I'd trust with my reputation. But this one was entirely different. Honestly, it scared me a little. The fidelity to my own speech and mannerisms was just that good. To put the platform through its paces, we supplied several hours of recordings taken from this very podcast to train an AI clone of my speaking voice. Four hours later, my voice clone was born. The side by side test recordings were spooky in their fidelity. I compared live recordings of myself reading a script against the AI, generated versions using identical texts, and even knowing which was which. I had to listen several times to tell the difference. Next came the real test, producing an entirely new episode. Episode 17, using an AI voice clone. The first AI pass was maybe 90% there. The fidelity was excellent, but the pacing and phrasing needed serious tweaks to be production ready. Turns out that punctuation and text formatting made all the difference. With some adjustments, the AI delivered a finished file that I felt confident enough to put head to head with my own live voice. Our production team handled it just like any other episode. The result? A workflow that produced high quality, authentic sounding spoken word content in my voice that is almost indistinguishable from me sitting down in front of a mic. It was a game changer. So now the elephant in the room is just because we can do this, should we? Podcasting is built on a social contract between host and audience. You extend me trust that what you're hearing is real, authentic and personal. If I break that trust, even unintentionally, the whole enterprise falls apart. That means disclosure is a non negotiable. Whenever and wherever AI is used, you must say so. If there's a reason for it, like efficiency, accessibility, or consistency, you need to share that too. Otherwise the audience won't just distrust the tool, they'll distrust you. For me, the authenticity lies in this. While the voice may be cloned, the stories and ideas are 100% mine. I'm not outsourcing thought leadership. I'm not faking a Persona. I'm simply using a new tool to make the heavy lift of recording lighter. For this episode, we disclosed it in three ways. One, a disclaimer in the show notes. Two, a disclaimer in the audio itself, and three, a contextual recap explaining why we chose to experiment at all. Transparency matters. But here's where things get tricky. What if my cloned voice gets misused? What if someone generates a deep fake of me saying things I'd never say? Or what if a former employer kept using my voice clone like long after I'd moved on without consent or compensation? These are not abstract risks We've seen likeness and voice disputes play out in Hollywood for decades. Crispin Glover famously sued Universal after a lookalike actor stood in for him in Back to the Future 2. Today, AI is magnifying that risk, outpacing current law around likeness, biometrics, and intellectual property. That's why we need policies and safeguards, both organizationally and legally. First, consent. No one's voice or likeness should be cloned without explicit informed consent. Second, transparency Creators must disclose when AI is used, why, and how. Third, ownership contracts should clearly define who owns the voice model, how it may be used, and what happens if the creator leaves. Fourth, security trained models must be safeguarded against theft or misuse, just like any other sensitive biometric data. And finally, expiration voice rights shouldn't live in perpetuity. There should be clear limits and renewals, just like any other intellectual property. Done right, voice cloning can remove barriers and enhance the creative process. But done wrong, it undermines trust, damages relationships, and erodes the very foundation of leadership communication. The promise of voice cloning is powerful. It lets creators like me deliver high quality content more efficiently and consistently. And it helps me show up for my audience, even when travel or logistics would otherwise prevent it. But that promise only matters if it is paired with honesty, consent, and safeguards. Trust is the true currency of leadership. Spend it wisely, because once it's gone, it's almost impossible to earn back. But if applied authentically and transparently, tools like voice cloning can help forge a truly lasting connection between host and audience. Thanks for listening. I'll see you soon.