Voice cloning is the reason most marketers try ElevenLabs and the feature most of them misunderstand. There are two versions. One takes ninety seconds of audio and gives you a passable copy in seconds. The other takes an hour of your audio, half a day of training, and gives you a voice your clients can't tell from the real thing. They live on different plans and they answer different questions.
We use ElevenLabs at our agency and we're affiliates. Every requirement below comes from the ElevenLabs voice cloning documentation and the ElevenLabs pricing page, checked on September 10, 2026. For the whole tool, read our ElevenLabs review.
Instant vs Professional at a Glance
| Item | Instant Voice Cloning | Professional Voice Cloning |
|---|---|---|
| Audio needed | About 1 to 2 minutes, no more than 3 | 30 to 180 minutes |
| Training time | Seconds | 3 to 6 hours, up to 24 |
| Whose voice | Audio you have the rights to | Your own voice only, verified |
| Starting plan | Starter, $6 a month | Creator, $22 a month |
| Slots | Counts against custom voice slots, 3 on Starter and 30 on Creator | 1 on Creator and Pro, 3 on Scale, 10 on Business |
| Languages | Follows the speech model | 39 listed for Professional |
| Result | Recognizably you, with a generic finish | Close enough for client-facing narration |
The honest summary is that Instant tells you what your voice sounds like through the model, and Professional is what you ship. We ran Instant for a week of internal drafts and it was fine for that. The moment a voiceover was going in front of a client, it was Professional or a human read.
Instant Voice Cloning, What It Needs
The Instant Voice Cloning guide is short and specific.
- About 1 to 2 minutes of audio. Clear, no reverb, no background noise, no music bed.
- Don't go past 3 minutes. ElevenLabs says the extra audio adds little and can make the clone worse.
- MP3 at 192 kbps or higher. 128 kbps is the floor. WAV usually doesn't help.
- File count doesn't matter. Total clean run time does. One good two-minute file beats six noisy clips.
The mistake we see is people feeding it a podcast excerpt with an intro sting and a co-host. The clone picks up the room and the crosstalk. Record two minutes into a phone in a closet and you'll get a better result than from an hour of edited show audio.
Professional Voice Cloning, What It Needs
- 30 to 180 minutes of good audio. Same format rules as Instant. More clean audio in that range produces a better clone.
- Voice verification. You read a prompt so ElevenLabs can confirm the voice is yours. Fail it and you wait 24 hours to retry.
- 3 to 6 hours of fine-tuning. Sometimes longer, up to 24 hours. Plan the recording for a day when you don't need the voice.
- Creator plan or above. $22 a month, one Professional slot. The Creator plan post covers what else comes with it.
Record the way you'll use it. If the clone is for ad reads, record ad reads. If it's for long narration, record narration. The model learns your pacing and energy from the sample, and a conversational hour produces a conversational clone.
The Rule That Matters, Your Voice Only
ElevenLabs is explicit in its documentation. "You can only create a Professional Voice Clone of your own voice. Even with their consent, you cannot clone someone else's voice." Verification exists to enforce it. That closes the door on the use case agencies ask us about most, cloning a founder or a client spokesperson so the team can generate their reads. It doesn't work on Professional, and the terms bar it.
- No client voices on Professional. The founder has to record, verify, and own the account seat.
- Rights apply to Instant too. Section 4(g) of the terms bars output you don't have the rights to. A signed release from the speaker is the minimum.
- Commercial use starts at Starter. Nothing generated on the free plan can be sold, cloned voice or not. The free plan post covers that.
How to Make a Professional Clone
- Upgrade to Creator or above. The pricing page lists Creator's first month at $11.
- Record 30 to 180 minutes in a quiet room, in the style you'll generate. Export as MP3 at 192 kbps or higher.
- Open Voices, choose Professional Voice Cloning, and upload the files. File count doesn't matter, total run time does.
- Complete voice verification by reading the on-screen prompt in your own voice.
- Wait 3 to 6 hours for fine-tuning. Check back within 24 hours if it runs long.
- Generate a real script, listen for the names and numbers it stumbles on, and re-record if a sound is consistently off.
Verdict
Instant is for finding out, Professional is for shipping. Two minutes on Starter tells you if the model likes your voice. An hour on Creator gives you one you can put in front of clients.
You can only clone yourself. If the plan was to clone a client, stop here. ElevenLabs bars it even with consent.
Audio quality beats length. A clean phone recording in a closet outperforms edited show audio every time.