Is this my real voice — or a cloned one?
We old timers remember the old Memorex commercial that asked, “Is it live or is it Memorex?” Well, my latest video is something like that. Some sections of the video use my real voice while other sections use a cloned AI voice. I want to know if you tell the difference?
In this post, I’ll walk you through exactly how I created my AI talking avatar. This is not a tutorial, but a high-level explanation of the process I used, including what I did wrong and what I learned.
If you’ve been curious about creating an AI avatar with voice cloning, this will give you a practical overview.
Why I Decided to Create an AI Talking Avatar
The reason is simple: I don’t enjoy being on camera.
With a traditional YouTube setup, you have to think about:
- How you look
- Lighting
- Hair
- Facial expressions
- Random things like food in your teeth
With an avatar, none of that matters.
The avatar handles the on-camera presence while I focus on writing. I much prefer writing to sitting in front of a camera. For someone who prefers creating behind the scenes, this changes everything.
The Tools I Used
I used HeyGen for the avatar video and Fish Audio for voice cloning.
Why two tools? Well, because I didn’t like the built-in voices inside HeyGen. Many sounded too high-pitched or generic. I wanted something with a deeper voice that actually sounded like me.
The Mistake I Made (So You Don’t Repeat It)
I immediately went to HeyGen to create the talking avatar. I quickly found out that that was backwards. The avatar needed a voice, and if the voice doesn’t sound right, the avatar doesn’t matter.
The voice carries the personality. It carries trust. It carries the connection. So I retreated and started again.
Lesson learned: Voice first, avatar second.
How I Cloned My Voice
Inside Fish Audio, I uploaded a short voice sample. It only took 10 seconds. Technically, it worked, but it didn’t fully sound like me. So, I uploaded a longer sample. That sample was about 90 seconds, and it changed everything.
The longer recording captured my natural pacing, rhythm, tone shifts, and speech patterns. Now, don’t get me wrong, it wasn’t a 100% match, but it was something I was willing to work with.
If you’re experimenting with AI voice cloning, don’t rush this part. Clean, longer samples produce noticeably better results.
Turning the Cloned Voice into a Talking Avatar
Once I had a voice I liked, I typed in my script. Fish Audio converted the text into speech usingmy cloned voice. I then downloaded the audio file.
Now that I had an audio file that sounded more like me. Went back to HeyGen.
Creating the Video in HeyGen
When I got to HeyGen, I selected the avatar I wanted to use, uploaded the audio file, and clicked “Generate”. Then I walked away.
Processing the file takes a few minutes; rather than sitting and looking at the computer screen, I got up and made a cup of coffee. HeyGen sends an email when the video is ready. As a matter of fact, the video was ready before I finished brewing my coffee.
It doesn’t get much easier than that. No filming. No lighting setup. No retakes.
Here’s the video
What Still Needs Improvement
Here’s the honest part.
While voice cloning can sound impressively accurate, emotional expression is harder. The pacing can feel slightly off. The emotional emphasis isn’t always natural.
Fish Audio does offer tools to improve emotional variation, I just haven’t mastered them yet. And that’s okay.
I’m not trying to perfect everything immediately. I’d rather improve incrementally than overwhelm myself trying to master every feature at once.
What I’ve Learned So Far
After playing around with creating talking avatars, here are some things I’ve noticed.
- Short, natural sentences work best. Long, complicated sentences don’t translate well into AI speech.
- Punctuation matters. Clear punctuation improves pacing and rhythm.
- Clean audio is important. Garbage in, garbage out. Background noise reduces cloning quality.
- Take it slow. Both platforms have a lot of features. Trying to become an expert in everything immediately is overwhelming.
- I’m striving for progress, not necessarily perfection
Who Should Try AI Avatars and Voice Cloning?
Honestly? Most people should experiment with creating a talking avatar. Both HeyGen and Fish Audio offer free tiers.
If you don’t like being on camera, prefer writing over speaking, and are curious about AI tools, it’s worth giving it a try. Especially if you’ve been holding back from video content because you don’t want to be on screen.
Is It Live… or Is It AI?
There were a few glitches in my video. That’s part of the learning process. But I’ll ask you the same question I asked at the beginning. Is it Live or is it Memorex? Could you tell which sections were cloned and which were my real voice?
Like it or not, artificial intelligence is here to stay. Rather than run from it or stick my head in the sand, this grandma has decided to learn about it and find out how it can help me in my everyday life.
The key isn’t avoiding the technology. It’s learning how to use it well.