01What VoiceStudio Is and How It Works With Claude Code
VoiceStudio is a free desktop app that does what the big paid voice websites do, but on your own computer. In this episode, Julian Goldie connects it to Claude Code so every video, ad and training lesson can go out in his voice without him recording a word.
VoiceStudio is open source, which means the code is public and free to use. It is one of the most popular projects on GitHub right now. Here is what it can do:
- Clone a voice from a short recording.
- Design a brand new voice from a written description.
- Read out long text.
- Dub a video into another language.
- Type out what you say.
The part that matters most is that VoiceStudio comes with a built-in connection for AI agents called MCP. Think of MCP as a plug. Once Claude Code is plugged in, Claude can make speech, clone voices and read audio just by you asking in normal words.
You do not have to learn the app. You tell Claude what you want, and Claude drives the app for you.
02Main Features of VoiceStudio You Can Use
Once Claude Code is connected, it can see a set of VoiceStudio tools. These are the jobs Claude can run for you just by asking:
- Generate speech from any text.
- Clone a voice from a sample.
- Design a voice from a description.
- Transcribe audio into text.
- List your voices so you can pick one.
On Julian's Mac, the app ran on the Mac's own graphics chip, which makes it fast on the newer Apple Macs. VoiceStudio also has a job queue, so you can line up a big batch and let it run while you sleep.
There are a few more features worth knowing. You can tie each AI agent to one saved voice, so whenever Claude Code speaks, it uses your voice twin by default. It can watch a folder and process anything you drop into it.
Every voice model shows its license so you can check before you install. And VoiceStudio adds an invisible watermark to the audio it makes, switched on by default, so the audio can be identified as AI made later.
Because it is free and runs on your own computer, there is no cap on how much audio you make.
03How to Set Up VoiceStudio With Claude Code
Julian opened Claude Code and sent one message asking it to install VoiceStudio on his Mac, connect itself to it and get it ready to clone his voice. Here is what Claude did:
- Installed the app with Homebrew. Homebrew is like an app store for tools.
- Found the connection settings. VoiceStudio has a page called Integrations that gives you a ready-made file for Claude Code.
- Added that file. From then on, Claude could see the VoiceStudio tools.
- Checked the app was healthy and running on the Mac's graphics chip.
- Picked a voice model. This is the brain that actually makes the speech.
That last step is where Claude did something smart. It read the license on the default model and stopped, because that model is free for personal use only, not for business. Instead of clicking accept, it told Julian.
Then it picked a different model called VoxCPM2. That model has a license that allows business use, it can clone a voice from a short sample, and it speaks 30 languages. This is a good reminder to always check the license of any voice model you pick.
04How to Clone Your Voice the Right Way
Most people think the more audio you give, the better the clone. What actually matters is clean audio: one clear recording of you talking with no music, no echo and no other voices.
If you record a fresh sample, follow these tips from the episode:
- Sit somewhere quiet with soft things around you, like curtains or a sofa, because bare walls make echo.
- Keep the mic about a hand's width from your mouth.
- Talk the way you normally talk to a customer, not the way you read a script.
- Say a mix of things: a question, a short sentence and a longer explanation, so the clone hears your full range.
Julian did not record anything new, because he already had plenty of his own audio. He pointed Claude at one of his videos and told it the first 30 seconds had music. Claude skipped the music, found 20 clean seconds of him talking, cut out the pauses, set the volume and typed out what he said.
Now he can ask for any line at any length. Read this email out loud in my voice. Make a 30-second welcome for new members.
Turn these 10 tips into 10 short voice notes. Claude writes the words, VoiceStudio speaks them, and the files land in a folder.
05Turning a Blog Post Into an Audiobook With VoiceStudio
A lot of people would rather listen than read, like when they are driving, walking the dog, at the gym or cooking dinner. If your content only exists as text, those people never get it. So Julian built an audiobook press.
He gave Claude the address of one of his blog posts and asked it to turn the post into an audiobook in his cloned voice, with a chapter for each heading. Claude did it in five steps:
- Read the page and pulled out just the article, with no menus, buttons or footer.
- Cleaned the text for listening, rewriting things that sound bad out loud, like lists of links, tables or web addresses.
- Split the post into chapters, one per heading.
- Sent each chapter to VoiceStudio in Julian's voice.
- Joined the chapters into one audio file, leveled the volume and saved it.
The post he used was about Hermes Agent for marketing. Because it is a system, it works on the next post too, and you can point it at a whole folder of posts.
One extra tip: ask Claude to add a short intro and outro in your voice, saying this is the audio version and the full article is on your website.
06Dubbing a Video Into New Languages in Your Own Voice
The third system is the dub desk. Recording everything again in Spanish, French or German is a huge job, so Julian asked Claude to do it. He gave it a short video of himself explaining an idea and asked for it in Spanish, French and German in his own voice.
Here is how it worked:
- VoiceStudio typed out what he said, with the timing of every line.
- Claude translated each line and kept it short so the new language fits the same time as the original.
- VoiceStudio spoke each line in his cloned voice in the new language.
- The new voice track went back on the video.
Step two is the tricky part. German, for example, often takes more words to say the same thing, so a straight translation runs too long. Claude tightens the wording so the speech lines up with the picture.
The result was the same video in the same voice in three new languages. For a business, that means:
- Your best video can reach people who would never have understood it.
- Your ads can run in new countries.
- Your onboarding can greet a new customer in their own language.
07Three Upgrades and One Mistake to Avoid
Once the three systems are running, Julian suggests three upgrades:
- Give Claude its own voice setting. Tie Claude Code to your saved voice twin, and every audio file it makes comes out in your voice by default.
- Design a second voice. Type a description, like a calm, warm female narrator in her 30s with a British accent, and VoiceStudio builds it. There is no recording and no real person behind it. Use it for an ad voice that answers yours, a story narrator or a friendly phone line voice.
- Use a watch folder. VoiceStudio can watch a folder and process anything you drop in. Save a new blog post to one folder and the audio shows up in another, without opening the app or asking Claude.
The one mistake to avoid is sending raw text straight off the page. Always have Claude clean the words for listening first: short sentences, no links, no tables, and numbers written the way you would say them out loud. The voice can only sound as good as the words you hand it.
And if a line sounds off, ask Claude to make it again. It is free to run, so you can make 10 versions and keep the best one.
08Staying Safe and Who VoiceStudio Is For
There is one rule to follow before you clone any voice. Do not clone anyone else without their permission, and do not use a cloned voice to pretend to be someone. VoiceStudio says the same thing on its own page.
Use it to scale your own voice. When you post a video with an AI voice, tick the box that says it contains AI-made content. And check the license of the voice model you pick.
Julian compares the old way and the new way. The old way: you record every piece of audio, redo it when you stumble, blog posts stay as text, and new languages mean hiring someone. The new way: you record one clean sample once, every post becomes an audiobook, and every video goes out in three languages in your voice.
Here is who can use it this week:
- Agencies can send each client a short voice update every Friday.
- Course sellers can add audio and translated versions of every lesson.
- Content creators can publish each article as an audio episode the same day.
- Local businesses can make a phone greeting and short how-to voice notes.
- Teams can turn process documents into spoken walkthroughs for new hires.