AI voices: how to create narration, dubbing and clone your own voice (and what not to do)
ElevenLabs and other AI voice tools: how they work, what they cost, free vs commercial plans, and the rules to stay out of trouble with other people's voices.
Synthetic voices no longer sound robotic. Today an AI can narrate a video with emotion, read an audiobook or dub a video into another language while keeping your voice. It’s a fantastic tool for creators, but it comes with very clear rules.
What you can do
- 🎙️ Narrate videos, shorts, podcasts and presentations.
- 📖 Read long texts aloud: audiobooks, articles or notes.
- 🌍 Dub a video into another language.
- 🧬 Clone your own voice so you don’t have to record every time.
- 🎭 Design a brand-new voice by describing how it should sound.
The tools
ElevenLabs is the industry benchmark for natural-sounding voices and a library of thousands of voices. You’ll also find AI voices in Google (inside Gemini and NotebookLM), Microsoft Azure, editors like CapCut, and open-source models that run on your own computer.
Free vs paid: the fine print
This is the part most people miss:
| Free plan | Paid plan | |
|---|---|---|
| Use | Personal, testing | Personal and commercial |
| Can I monetize? | Usually no | Yes, under the terms |
| Attribution | Sometimes required | Usually not |
On ElevenLabs, for example, the free plan doesn’t include a commercial license: if you’ll use the voice on a monetized channel or for a client, you need at least the Starter plan, a few dollars a month. Always check the current terms on their site.
Voice cloning: the golden rules
- Only your own voice, or someone who’s given you written permission. Cloning a celebrity’s, a voice actor’s or an acquaintance’s voice without permission can violate their rights and is prohibited by the platforms themselves.
- Never to deceive. Using a cloned voice for scams, like fake calls from a relative asking for money, is a crime.
- Disclose that it’s AI. YouTube requires labeling realistic synthetic content, and the EU AI Act requires artificially generated content to be identified. We do it in our own videos: “AI-generated voice.”
How to get natural narration
- Write for the ear: short sentences, clear punctuation.
- Spell out numbers and acronyms the way they’re said if the voice trips over them.
- Play with stability: lower stability gives more emotion; higher gives a steadier read.
- Generate sentence by sentence for the key parts and pick the best take.
Beware of voice scams
The flip side: scammers already use cloned voices to impersonate family members. Agree on a family code word and, for any urgent call asking for money, hang up and call back on the usual number.
Looking for music to go with your narration? Read making music with AI.
