100+ agencies trust Arcytex 12x HighLevel Award winners 14-day refund guarantee 1,000+ HighLevel builds 50+ SaaS scaled past 100 users 100+ agencies trust Arcytex 12x HighLevel Award winners 14-day refund guarantee 1,000+ HighLevel builds 50+ SaaS scaled past 100 users

July 5, 2026

AI Voice Calls in Your Own Cloned Voice

AI voice calls can now sound exactly like you. Here's how voice cloning works in GoHighLevel Voice AI — the 3-minute script, the queue, and why the cloned voice is only the last 5% of a real AI receptionist.

By Arcytex · GoHighLevel, AI Automation, Voice AI, AI Receptionist, Voice Cloning

AI voice calls can now go out in your own voice. GoHighLevel Voice AI lets you clone your voice from one short recording, so the agent that answers the phone or dials your leads sounds exactly like you — not a stock robot. It takes a three-minute script, a few minutes in the queue, and an agent that’s already built to run it.

This post distills Nuno Tavares’s video, “GoHighLevel Voice AI: Clone Your Voice for AI Calls (Full Setup).”

What does it mean to clone your voice for AI voice calls?

Voice cloning copies the sound of your voice from a sample recording, then uses that clone to speak whatever text the AI generates. In GoHighLevel Voice AI, that means your AI voice calls — the inbound receptionist and the outbound follow-up — play in your voice instead of a generic one.

Here’s the diff. A stock AI voice is fine, but it’s obviously not you. A cloned voice makes the agent feel like your business answered, because in a way it did. The clone gets recorded, queued, processed, and then attached to an agent you’ve already set up.

That last part matters. Cloning your voice does not build the agent. It skins one you already have.

How do you clone your voice in GoHighLevel Voice AI?

The whole thing is a short recording session inside an existing voice agent. Build the agent first, then record the voice and attach it. Five steps.

Step 1: Have an agent already built

You can’t clone a voice into thin air — you attach it to a voice AI agent that already exists. If you haven’t built one yet, that’s the real work, and it comes first. The cloning below assumes you’ve got an agent ready to wear the new voice.

Step 2: Open the voice settings

Go to AI Agents → Voice AI → Agent List and open the agent you want to give your voice to. Hit Edit, then click into the voice area for that agent. Click My Voices, and you’ll see an Add a Custom Voice button.

Step 3: Choose record or import

You get two paths. If you already cloned your voice somewhere like ElevenLabs, import it. If not, record right here in GoHighLevel using your own mic. Either way, you name the voice and pick a provider (ElevenLabs or Cortista in the video).

Step 4: Record the script

Hit record and read your script straight through. Aim for about three minutes of clean audio, which lands around four to five megabytes — comfortably under the 10 MB cap the recorder allows. Read at a normal pace, in your normal tone. This sample is the only thing the clone learns from.

Step 5: Save, then wait for the queue

Save the recording, select your new voice, and hit Save Voice. You’ll see a message that the voice has been queued and will appear shortly. It’s not instant — give it a few minutes to process, then start a test web call and hear it back.

In the video, the finished clone passed the test on the first try. The agent answered a web call and it sounded, in Nuno’s words, “completely my voice.”

“That is completely my voice… not for nothing, but that’s pretty cool.”

What should the voice-cloning script say?

The script is the whole ballgame, because the clone is only as good as the three minutes you feed it. A flat, monotone reading gives you a flat, monotone clone.

Nuno’s move here is worth stealing: he had an AI write the script for him first, then read it off the screen. A good clone script does a few things on purpose:

  • Covers a range of sounds. Mix short and long sentences, questions and statements, so the clone hears your voice do more than one thing.
  • Includes the words your agent will actually say. Numbers, dates, times, appointment language — the stuff a receptionist or follow-up bot says every day. If the agent will read back phone numbers and booking times, put numbers and times in the sample.
  • Sounds like you talk. Read it out loud before you record. If a line feels stiff in your mouth, rewrite it. The goal is your natural speaking voice, captured clean.

Record in a quiet room. Background noise gets cloned too.

Why is a cloned voice only the last 5% of an AI receptionist?

Here’s the honest gotcha. Cloning your voice is the flashy, easy part — a three-minute recording and a few minutes in a queue. It changes how the agent sounds. It changes nothing about whether the agent actually works.

Whether your AI voice calls book the appointment, answer the real question, pull the right info from your CRM, and hand off cleanly to a human when they should — none of that comes from the voice. It comes from the agent built behind the voice. The voice is the last 5%. The system is the other 95%.

We say this to clients across the table all the time: a beautiful clone on a badly built agent just means your business now sounds exactly like you while it drops the ball. Get the system right first. The voice is the finishing touch, not the foundation.

That’s the part we build. AI that answers the phone — inbound receptionist, outbound follow-up, database reactivation — wired to your pipeline and your actual call flow, with the voice cloned on top once the agent behind it works. We’re 12x HighLevel Award winners with 1,000+ builds shipped, backed by four written guarantees: zero downtime, on-time delivery, a 14-day money-back window, and no long-term contract.

Frequently asked questions

How long does voice cloning take in GoHighLevel?

Not long, but it’s not instant either. After you save the voice, it gets queued and processed, and appears in your voices “shortly” — a few minutes in the video. Wait for it to finish, then run a test web call before you point real traffic at it.

How long should the recording be?

About three minutes. That’s roughly four to five megabytes of audio, which sits comfortably under the 10 MB file cap the in-app recorder allows. Long enough to capture the range of your voice, short enough that you can read it in one clean take.

Can I use a voice I already cloned in ElevenLabs?

Yes. When you add a custom voice, you can import from ElevenLabs instead of recording in GoHighLevel. If you’ve already built and tuned a clone there, import it and skip the recording step.

Do I need to build the agent before cloning my voice?

Yes. The cloned voice attaches to a voice AI agent that already exists — it doesn’t create one. Building the agent that handles the actual conversation is the real work and comes first. If you’d rather not build it yourself, we install a team of AI agents that do the job before you ever record a voice.

Does a cloned voice make the AI better at its job?

No. A clone changes how the agent sounds, not how well it books appointments, answers questions, or reads your CRM. Those depend entirely on how the agent is built. Treat the voice as the finish, not the fix.

Get it built right the first time

Cloning your voice is a three-minute recording. Building an AI receptionist that actually answers every call, books the appointment, and follows up on dead leads — that’s the system underneath it. Do that part wrong and a perfect clone just means your business sounds like you while it loses the lead.

That’s the work we do every day. Want AI voice calls that sound like you and actually close? Book a scope call and we’ll map it to your setup.

Liked this post?

Ready to ship something real?

30 min · No pitch · Written plan either way.

  • 14-day refund
  • No contracts
  • Cancel anytime
  • 12x HighLevel Awards