Key takeaways
- We cloned my voice, on purpose, for exactly one job: the daily ad. Every YouTube video still has me talking, live, on camera.
- The daily ad is the only place the clone shows up. That's the line, and we drew it after a different generation went wrong.
- We do not regenerate my face anymore. A cloned voice over a real background is honest. A generated face pretending to be me is not, and we found that out the hard way.
- The same ad script rotates across four verticals a week so the same feature never gets pitched the same way twice.
- None of it posts without me reading it first. The voice is automated. The yes is not.
Somewhere out there is a version of my voice that has never had a bad day, never lost its place mid-sentence, and can record a 30-second ad at 6am without coffee. I approved it on purpose, and I still think about that sentence more than I expected to.
Here's the actual story, because "AI voice clone" sounds like a bigger leap than what we actually did.
Where the line got drawn
Before the voice clone, we tried something bigger: a fully generated version of me, face and all, for a webinar video. Six scenes, each one generated separately. When we pulled a still frame from the middle of each scene to check it, we found four visibly different men. Different hairlines, different jaws, different faces entirely. Each generation had reinterpreted "me" from scratch, with nothing holding the face consistent between scenes.
That killed the idea of a generated face, full stop. Not paused, not scaled back. Dead, and I'm the one who films anything that needs my actual face on screen now.
But the voice clone survived that same review, because a voice and a face are not the same risk. My voice doesn't have a jawline that can drift between takes. It's trained on hours of me actually talking, and when it's laid under a real, unaltered background clip instead of a fake version of my body, nobody is being asked to believe they're looking at a generated person. They're hearing me, over footage that never pretended to be a live recording of me saying those exact words. That distinction is the whole reason one survived and the other didn't.
What the daily ad actually is
Every day, one ad gets built: my voice clone reading a script about a specific automation, over a branded motion-graphic background, with captions burned in so it plays fine muted. No face, because there's no face to fake. Just voice, visuals, and a link.
The script rotates through four verticals across the week: home services, real estate, financial services, and agencies. Same rule every time, no repeats: never pitch the same feature twice. One day it's speed-to-lead auto-texting a website form. Another day it's automated appointment reminders cutting no-shows. Another it's pipeline nurture so a not-ready-yet lead doesn't go cold. The point isn't to hammer one pitch, it's to show the range of what actually gets built.
Why the voice still needed a human decision
The clone didn't get approved by default just because the technology existed. It got approved once, for this specific use, after I looked at what it would replace and decided a cloned voice over honest visuals was a fair trade for a tool I couldn't otherwise run daily by myself. YouTube videos are a different decision entirely. Those still get recorded live, on a timed script, because that's the channel where people are getting to know the actual me, not a daily ad selling a system.
That's really the whole model. Not "AI can do this so let it," but "here's specifically what I'm comfortable automating, here's specifically what stays human, and here's why the two aren't the same call."
Why this matters for Bay Area and San Jose businesses
Most local business owners I talk to either avoid AI voice and video entirely because it feels dishonest, or they use it everywhere without thinking about where the line should sit. Neither is the right call. The tool is fine. The question is always what you're using it to replace, and whether the replacement is honest about what it is. A cloned voice over real footage, clearly repeated daily and clearly not trying to be a live take, is a different thing than a generated face standing in for a real conversation.
Practical steps
- Before automating any part of your own presence, ask what specifically breaks if it goes wrong. A voice reading a script breaks differently than a face pretending to say something you never said.
- Draw the line in writing once, and keep it. Ours is: my real voice and face for anything long-form, the clone for one daily short.
- Rotate what you actually pitch. Repeating the same hook daily wears out an audience faster than posting daily ever will.
- Keep a human approval step on anything that goes out with your name or voice on it, automated or not.
We build systems like this at On Point Tech Solutions in Los Gatos, serving San Jose and the wider Bay Area. Every one, you own outright, done for you or taught with a full SOP. If you're weighing where AI should and shouldn't stand in for you, let's talk it through. Start at optechsol.llc.