VoiceStudio

1. What is VoiceStudio?

VoiceStudio is a desktop app for AI voice work that runs on your own computer. It does six main things: voice cloning (give it a sample recording, then have that voice read any text), designing a new voice from a description, dubbing videos, dictation, turning audio into text, and producing audiobooks. The project describes itself as an open-source alternative to ElevenLabs and lists support for 646 languages.

The app is free and open source (AGPL-3.0 license). Local work runs on your hardware and needs no account. The app includes a page comparing Free and Pro; see https://voicestudio.sh for Pro pricing and details.

It ships as a ready-made installer for Windows, macOS and Linux, so non-technical users can install it.

A good fit if:

• You make videos, podcasts or audiobooks and want to generate voice-over on your own machine.
• You do not want to pay a monthly fee for an online voice service.
• Your computer has 8 GB of RAM or more; an NVIDIA graphics card or an Apple Silicon Mac makes it much faster.

Not a good fit if:

• You only have a phone.
• You have an Intel Mac: the interface opens, but local AI processing cannot run on those machines.
• You plan to clone someone's voice without their consent. The project asks you to clone voices only with permission.

2. What do you need?

Item	Details
Operating system	64-bit Windows, macOS (Apple Silicon; Intel Macs get the interface only, paired with a remote machine), or 64-bit Linux.
RAM	The setup check uses 8 GB as the minimum. 16 GB is more comfortable for long video dubs.
Graphics card	Optional. On Windows only NVIDIA cards are used for acceleration; AMD and Intel graphics run on the CPU, which works but is slower. On Apple Silicon Macs acceleration is automatic.
Disk space	The installer is a little over 200 MB. The app then downloads a Python environment (a few GB) and voice models (a few GB each). The docs for the older installer say about 10 GB free; treat that as a minimum.
Internet	Needed for the installer, the runtime and the models on first use.
Hugging Face account	Optional. Hugging Face is a hosting site for AI models. A free access token speeds up model downloads, and some larger models require one.

You do not need to install Python, FFmpeg or any developer tools yourself. The app handles that on first launch.

3. Step-by-step installation

Step 1: Download the right installer

Open the official releases page: https://github.com/debpalash/VoiceStudio/releases/latest

Scroll to "Assets" and pick the file for your machine. The latest version when this guide was written is v0.5.6 (released 23 September 2026):

Your machine	File to download
Windows	VoiceStudio-Electron-0.5.6-win-x64.exe
Apple Silicon Mac (M1 or newer)	VoiceStudio-Electron-0.5.6-mac-arm64.dmg
Intel Mac	VoiceStudio-Electron-0.5.6-mac-x64.dmg (interface only)
Linux	VoiceStudio-Electron-0.5.6-linux-x64.AppImage or the .deb file

What success looks like: the downloaded file name starts with VoiceStudio-Electron-. Only take files with "Electron" in the name: the older Tauri-based app was retired after v0.5.3.

Step 2: Run the installer

Windows: double-click the .exe and follow the setup wizard.

macOS: double-click the .dmg and drag VoiceStudio into the Applications folder.

Linux: install the .deb with your package manager, or make the .AppImage executable and open it.

Note: The project's release notes say the installers are unsigned or ad-hoc signed and not notarized by Apple, so Windows and macOS may show a trust warning. Section 5 explains what to do. Only proceed past the warning if you downloaded the file from the releases page in Step 1.

What success looks like: VoiceStudio appears in your list of applications.

Step 3: First-launch setup

Open VoiceStudio. The setup screen checks your machine (RAM, disk, compute device). When asked, choose Install local runtime so the app downloads and builds its background engine on your computer.

What success looks like: the progress bar moves through each stage and then drops you into the main workspace. This can take many minutes depending on your connection, since several GB are downloaded.

Alternative: a one-line install

For people comfortable with a command line. The script downloads the latest Electron release and verifies its checksum before installing.

macOS / Linux:

curl -fsSL https://voicestudio.sh/install | sh

Windows (PowerShell):

irm https://voicestudio.sh/install | iex

If you use a coding assistant such as Claude Code, Codex or Cursor, the project README also provides a ready-made prompt that asks the assistant to install the app for you.

4. First use

The first thing the project suggests trying is voice cloning:

1. Open Voice cloning (under the Voice group in the sidebar).
2. Pick an existing voice or add a reference recording. The sample should be clean: one speaker, no music, no background noise.
3. Type the text you want that voice to read.
4. Click generate. If the app asks to install a model, agree; it shows the size before downloading.

What the result looks like: a playback bar with an export button appears below the text composer. Press play to listen.

Note: The first generation after opening the app is always the slowest, because the model has to load into memory. Judge the speed from the second generation onward.

Other areas to explore later: Design (describe a voice to create it), Dubbing (dub one video or a whole folder), Transcribe (audio to text), Stories and Audiobook.

5. Common problems and fixes

Before fixing individual errors, let the app check itself: Settings → About → Run self-check. It verifies the compute device, ffmpeg, disk space, data-folder permissions, RAM and access to the model hub, with a hint for anything that is off.

Symptom	Cause	Fix
macOS says the app cannot be opened because the developer cannot be verified	The installer is not notarized by Apple.	In Finder, right-click VoiceStudio → Open → Open again. On macOS 15: double-click once, then go to System Settings → Privacy & Security, scroll down and click Open Anyway. You only do this once.
macOS says the app "is damaged" with no Open option	The download was corrupted, or it is an old build.	Download the latest release again.
Windows warns about an unrecognized app when you run the installer	The installer is unsigned or ad-hoc signed (per the release notes).	Confirm you downloaded it from the project's releases page before allowing it. The release includes a SHA256SUMS.txt file for checking the download.
Setup shows "System RAM" in red and will not continue	The machine is below the 8 GB minimum.	Update to the latest release (it now accepts 8 GB machines that report slightly less). If the machine really is below the minimum, long dubs may run out of memory; the project's troubleshooting page describes how to bypass the check.
"Can't reach the local backend" appears while generating or dubbing	Not a network problem. The job is too heavy for the graphics memory or CPU and stalls.	In Model Catalogue, choose a smaller speech-to-text model. Use Flush in the top toolbar to free memory before dubbing. Test with a 10-second clip first.
Everything is very slow	The app is running on the CPU without you realizing, or free RAM is low.	Check Settings → Performance: the device line shows cuda, mps or cpu. Close browsers and other apps. On Windows, AMD and Intel graphics always run on the CPU.
Intel Mac: setup stops and downloads nothing	Local AI processing is not supported on Intel Macs.	Use the remote connection controls on the setup screen, or install on an Apple Silicon Mac, Windows or Linux.

When reporting a bug, use Settings → About → Save diagnostic bundle. Full troubleshooting page: https://github.com/debpalash/VoiceStudio/blob/main/docs/install/troubleshooting.md

6. Uninstalling and updating

Updating. Download the new installer from the releases page and install it over the existing app. Saved voices, projects and settings are kept. If setup asks, choose Install local runtime to refresh the engine. On macOS the project recommends updating manually with the installer.

If you are still on the old Tauri app (v0.5.3 or earlier): close it, back up its data folder, install the Electron app, confirm your voices and projects are present, and only then remove the old one. There is no automatic migration tool.

Uninstalling. Follow this order, because the uninstaller does not delete your data:

1. Remove the data first, from inside the app: Settings → Storage → Remove all data. It lists each folder with its real size, lets you choose whether to also delete the shared Hugging Face model cache, asks you to type DELETE, then removes everything and quits.
2. Remove the app. On macOS and Linux (script installs):

curl -fsSL https://voicestudio.sh/install | sh -s -- --uninstall

On Windows (PowerShell), this opens the uninstall wizard:

$env:VOICESTUDIO_INSTALL_MODE='uninstall'; irm https://voicestudio.sh/install.ps1 | iex

If you only want disk space back and want to keep the app, use Settings → Storage → Reset & remove to delete just the downloaded models or just the settings.

7. Frequently asked questions

Does it cost money? The app is free and open source. There is no per-character or per-minute fee. The app has a Pro page; see the official site for pricing.

Does it need internet? Yes on first use, to download the runtime and models. Local work then runs on your hardware. Features that connect to outside services are optional.

Are my voice and files sent anywhere? According to the project, local workflows run on your machine, with no accounts and no cloud state. Remote services are optional, and usage analytics are sent only if you consent.

Will it run on a weak computer with no graphics card? It runs on the CPU, but several times slower. A machine with less than 8 GB of RAM is blocked at the setup check.

Can I use the generated voices commercially? The app is AGPL-3.0, and each voice model has its own license. The project advises reading the model's license before commercial use, and cloning voices only with permission.