Recently, I cancelled my recurring ElevenLabs subscription and moved to pay as you go. I wanted a more cost-effective option, so I went looking for something that could run on a laptop I already own.

What I found

VoiceStudio, previously called OmniVoice Studio, is a free, open-source desktop app that runs voice work locally. I tried it on my MacBook Air with 24GB of RAM.

My first five minutes

I downloaded the Mac version from the VoiceStudio site. Installation was straightforward. The cloning workflow had three steps: Sample, Script, Speak.

For the sample, I recorded about 30 seconds of my own voice. For the script, I used my recent EdTechStuff post about old scheduled Copilot prompts retiring on 31 December 2026. It runs to roughly 400 words.

Then I selected generate. It took about 90 seconds. Within five minutes of installing the app, I had cloned my voice and created a spoken version of my own post. That’s fast enough to use. A tool that takes an afternoon to produce one clip stays in the “interesting” pile.

How it sounded

It sounded like me. Properly like me. Even my wife agreed. It was very realistic.

It had one problem: it struggled to say “Copilot”. Product names can trip up text-to-speech tools. Respelling the word phonetically in the script is the usual workaround. Either way, listen through before publishing.

The cost

I’m on ElevenLabs pay as you go. Its top-up credits last 12 months, which suits patchy use. The current Starter plan is listed at $6 a month, with a monthly credit allowance. Pay as you go avoids committing to that recurring subscription, although the available credits and prices can change. Check ElevenLabs’ current pricing and credit terms.

VoiceStudio’s desktop app is free and open source. Its current site lists Pro at $99 per user per year and a Lifetime option at $199, with a seven-day trial. The app’s licence and the selected speech model’s licence are separate. VoiceStudio says the OmniVoice pretrained model is non-commercial, so check the terms for the exact model you use before publishing audio or using it for paid work. VoiceStudio’s licence guide explains the distinction.

You pay in hardware and time instead. I ran my test on a 24GB MacBook Air and can only report how it performed on that machine. The free app works locally, though it may connect to download models or check for updates. Try it on your own device before planning around it. VoiceStudio’s privacy notice describes what stays local and when the app uses the network.

Consent

I cloned my own voice, which keeps this simple. VoiceStudio’s guidance says to clone only voices you have the right to use. That includes colleagues and governors.

The local setup helps, but in a school or trust I’d still expect a conversation with the data protection lead before anyone clones a staff member’s voice, even with their agreement.

The verdict

For spoken versions of my own posts, this did the job. It was quick to set up, sounded like me and added no software cost for this test. The Copilot pronunciation is a small fix I can live with.

It also looks useful for creating short speech extracts for my mini course generator. I’ve only scratched the surface of what this package can do, but it’s definitely got promise. I’ve tested one post on one laptop, using my own voice. For the question I started with, whether there’s a cheaper way to do this, the answer is yes.

Official references