AI Voice Cloning: I Cloned My Voice and It's Kinda Creepy
So last Friday night—yeah, Friday, not even during work hours—I spent two hours cloning my voice. Not because I needed to. Just because I could. And honestly? The result freaked me out a bit. My own voice reading back words I never said. Weird flex, I know.
But here's the thing. This tech isn't just a party trick anymore. People are using AI voice cloning for actual work. Audiobooks, YouTube videos, podcasts, even customer service bots. And some folks are making real money from it. Not "quit your job" money—let's be realistic—but enough to notice.
What Even Is Voice Cloning?
Okay, quick explanation. You record yourself talking for like 10-30 minutes (depends on the tool). The AI learns your voice patterns—pitch, tone, rhythm, all that. Then it can "speak" anything you type. In your voice. Scary? Yeah, kinda. Cool? Also yeah.
I used ElevenLabs for my test. Took me 15 minutes of recording random sentences from a book. Then boom. My digital twin could say whatever I wanted. First thing I made it say? "I'm a robot and I don't know it." My girlfriend did not find it as funny as I did.
Tools That Actually Work (I Tested Five)
I didn't just try one. That'd be lazy research. I tested five different platforms over two weeks. Some were great. Some were... not.
The ones worth your time:
- ElevenLabs: Best quality, hands down. Voices sound natural. Free tier gives you 10k characters monthly. Paid starts at $5/month. I've used this the most.
- Play.ht: Good for longer content. Slightly robotic on some words but overall solid. Free trial, then $31/month. Pricey but they have voice library access.
- Murf.ai: Easy interface. Great for beginners. Quality is... okay. Not bad, not exceptional. $19/month. I'd say it's the "safe middle" choice.
The two I wouldn't recommend? One sounded like a 1990s GPS. The other had this weird echo thing going on. Not naming them because maybe they've improved since I tested. (Or maybe they haven't. Who knows.)
Real Uses People Don't Talk About
Everyone focuses on audiobooks and YouTube. Sure, those work. But I've seen weirder applications that actually make sense:
Training videos for companies. My friend Sarah—works in HR—needed to create 40 onboarding videos. Recording herself 40 times? Nightmare. She scripted everything, used Murf, done in three days. Her boss loved it. She got a bonus. True story.
Language learning content. One guy I know creates pronunciation guides. He types phrases, the AI reads them in different accents (British, American, Australian). Sells them on Gumroad. Makes around $400/month. Not life-changing but it's passive income while he does his main job.
Voiceovers for social media. TikTok and Instagram Reels need audio. Recording yourself every single time gets old fast. Some creators batch-write scripts, generate voices, edit videos. Way faster workflow.
The Ethics Thing We Should Probably Discuss
Look, I can't write about voice cloning without mentioning this. The tech can be misused. Scams, impersonation, fake news—all possible. And that's... concerning.
Most legit platforms have rules. ElevenLabs requires you to confirm you have rights to the voice. They ban users who violate this. Play.ht has similar policies. But enforcement? That's the tricky part.
My take: use it for your own voice or with clear permission. Don't be a jerk. Simple rule.
⚠️ Important: Creating someone else's voice without permission is illegal in many places. Seriously, don't do it. Beyond legal issues, it's just wrong. If you're thinking "but what if..." — stop. The answer is no.
How to Start (Without Overthinking It)
You don't need fancy equipment. I used my phone's voice recorder. Yeah, really. Here's what worked for me:
Simple 4-step process:
- Find a quiet room. Like, actually quiet. Background noise messes with the AI. I used my closet. Felt dumb but it worked.
- Record 10-20 minutes of yourself reading. Anything works—book, article, grocery list. Just speak naturally. Don't do a "radio voice." Be you.
- Upload to your chosen platform. ElevenLabs processes it in like 2 minutes. Others take longer.
- Test with random sentences. Type something weird to see if it sounds right. I typed "Purple elephants prefer disco music on Tuesdays." If that sounds natural, you're good.
That's it. Seriously. People overcomplicate this. The tech does most of the work.
What I'd Do Differently Next Time
First attempt? I rushed the recording. Spoke too fast, stumbled over words. The AI picked up my mistakes. So the cloned voice had these little hesitations that sounded off.
Second attempt (different platform), I read slower. Clearer. Big improvement. The voice sounded smoother, more natural. Lesson learned: take your time on the source recording. Garbage in, garbage out, you know?
Also—and this surprised me—recording at different times of day made a difference. Morning voice vs evening voice. Stick to one or your clone might sound inconsistent. Found that out the hard way.
Cost Breakdown (Real Numbers)
Everyone asks about pricing. Here's what I actually paid:
- ElevenLabs free tier: $0 for 10k characters/month. I used this for testing. Enough for short projects.
- ElevenLabs paid: Started at $5/month for 30k characters. I upgraded to $22/month for 100k. Worth it if you're doing client work.
- Play.ht: $31/month. Canceled after two months. Good tool but I didn't need two subscriptions.
- Murf.ai: $19/month. Kept this one for simpler projects.
So realistically? Start free. If you're making money from it, $5-22/month is reasonable. Don't pay for premium features you won't use. (I did that. Waste of $40.)
Things That'll Trip You Up
Pronunciation of uncommon words. The AI guesses. Sometimes it's wrong. I had it butcher "açaí" three different ways before I respelled it phonetically as "ah-sigh-ee." Annoying but fixable.
Emotional range is limited. Happy, sad, excited—the AI tries but it's not perfect. You can adjust settings but it takes practice. Don't expect movie-quality voice acting right away.
Some platforms have commercial use restrictions on free tiers. Read the fine print. ElevenLabs terms spell this out clearly. Most require paid plans for selling content. Makes sense, just know it upfront.
FAQ (Questions I Get Asked)
Q: Can people tell it's AI?
A: Depends on the quality and length. Short clips (under 30 seconds) usually sound fine. Longer content? Some listeners notice small glitches—weird pauses, flat emotion. Getting better though. Like, way better than a year ago.
Q: Is it legal to monetize AI voices?
A: If it's your voice and you're on a paid plan that allows commercial use, yes. If you're cloning someone else or using free tier against ToS, nope. Check each platform's rules. They're all slightly different.
Q: How long does it take to create an audiobook?
A: I helped a friend do a 30k-word book. Took about 6 hours total—formatting the text, generating audio, editing out mistakes, exporting. Traditional recording would've been 30+ hours easily. Huge time saver.
Q: What about accents and languages?
A: English works best (obviously, that's what most tools optimize for). But ElevenLabs supports 29 languages now. Accents... hit or miss. American/British English are solid. Heavy regional accents might confuse the AI.
Q: Can I update my voice clone later?
A: Yeah. Just record new samples and retrain. Useful if your voice changes or you want to add more emotional range. I redid mine after getting better at recording technique.
Where This Tech Is Heading
Honestly? It's gonna get wild. Real-time voice changing during calls. Live podcast "guests" that are AI. Personalized audiobooks where characters sound like your family. Sounds sci-fi but it's coming faster than people think.
Some folks worry about job losses—voiceover artists, narrators. Valid concern. But I've also seen new jobs pop up: AI voice trainers, synthetic media editors, prompt engineers for audio. The landscape shifts. Always does.
My guess? Within two years, most YouTube videos will use some form of AI voice. Not all, but most. It's just more efficient. And quality keeps improving. We're early but not that early.
Final Thoughts From My Weird Experiment
Cloning my voice was 70% fascinating, 30% unsettling. Hearing "myself" read text I didn't speak out loud messes with your head a bit. You get used to it. Or maybe you don't. Still figuring that out.
But as a tool? Really useful. I've used my cloned voice for work presentations (when I'm sick and sound terrible), quick video demos, even birthday messages when I'm traveling. Practical applications outweigh the weirdness. For me, anyway.
If you're curious, just try it. Free tiers exist for a reason. Spend an hour. See what you think. Maybe you'll love it. Maybe you'll delete it immediately and never speak of this again. Both reactions are valid.
Try it yourself: Pick ElevenLabs or Murf.ai, record 10 minutes, see what happens. Then come back and tell me what you thought. Did it freak you out too? Or am I just weird?
Please when you post a comment on our website respect the noble words style