Cloning My Voice with AI – What Happened to Theo Von?

Cloning My Voice with AI (or at Least Trying To) I’ve been using an AI voice that sounds like Theo Von to narrate my videos. It’s been a fun setup — adds a bit of personality, and let’s be honest, sounds way better than my own English pronunciation.

But today, when I sat down to create a new video, I found out that the voice — my precious Theo Von — was no longer available on Play.ht.

I went to the site, opened one of my older video scripts, and the voice I had used before didn’t even have a name anymore. Just… blank. I tried to find it manually — searched through the voice list. Typed in “Theo Von” — nothing. Tried “Chris,” which is what the Theo-sounding voice used to be called. Still nothing.

So that’s it. Time to delete the channel. Maybe Theo Von found out and got mad I was using his AI twin.

What now?

Well, naturally, I turned to Mr. Robot himself — ChatGPT.

Here’s what it suggested:

First place: ElevenLabs. It’s great. But it’s also paid. So… moving on.

Second place: OpenVoice. Looks free. That’s all I needed to hear. Let’s give it a shot.

I’ve wanted to voice my content myself for a while now. But again, my English isn’t exactly native-level, and I’m not sure people would want to listen to me for 10 minutes straight. So the idea of cloning my voice — but letting AI clean it up — sounded perfect.

I quickly skimmed through the OpenVoice description and saw the magic words: multilingual voice cloning. That’s it. This is exactly what I need.

So I cloned the repo and opened up notebooks/demo_part1.ipynb.

First cell… boom. Problem. Missing models. No checkpoints included. Classic.

But no worries — after a bit of Googling, I found the right files, dropped them in, restarted everything, and ran it again.

This time, it seemed to work.

I recorded a short voice sample, re-ran the notebook… and finally got my AI-generated audio.

Let’s hear it.

Disappointment. Instead of a real clone of my voice, I got a very generic TTS voice with maybe a hint of my tone… but absolutely none of my personality.

The next step would’ve been training my own model — something much more complex than just feeding it a short sample.

Yeah… no thanks.

So I just picked a different pre-made voice. If you found this interesting — or have a better solution — drop a comment. Thanks for reading.