Home / Guides / Install VoiceStudio
How to Install VoiceStudio (Step-by-Step)
VoiceStudio is the open-source, fully-local ElevenLabs alternative with 77,800+ GitHub stars. Here's how to get it running on your machine in about 10 minutes.
Quick answer
Install VoiceStudio with one command, launch the desktop app, add a 10–30 second reference recording, and generate your first clip. The app runs locally; the first model download is the longest step. macOS, Windows, and Linux are supported, with NVIDIA CUDA, Apple Silicon Metal, and CPU-only modes.
What you can do with it
VoiceStudio bundles six workflows in one desktop app: voice cloning, voice design, video dubbing, dictation, transcription, and audiobook creation — across up to 646 languages depending on the engine. It runs entirely on your hardware, and exposes a local API and MCP server if you want coding agents to drive it. Local mode does not upload your recordings; optional remote services are opt-in.
Step 1 — Check your hardware
| Hardware | What you get |
|---|---|
| NVIDIA GPU (Windows / Linux) | CUDA acceleration — fastest option |
| Apple Silicon (M1–M4) | Metal (MPS) acceleration |
| No dedicated GPU | Fully usable on CPU, slower; needs ~5 GB free disk for the CPU build of PyTorch |
| Windows on ARM (Snapdragon X) | Experimental — native ARM64 app, CPU only |
| Intel Mac | App UI only; connect to a remote backend |
No GPU? Don't worry — CPU mode works fine for occasional use. Expect roughly real-time-or-slower generation instead of near-instant.
Step 2 — One-command install
macOS / Linux
curl -fsSL https://voicestudio.sh/install | shWindows (PowerShell)
irm https://voicestudio.sh/install | iexPrefer a manual download? Grab the latest release from the GitHub releases page — macOS .dmg, Windows .exe, Linux .AppImage or .deb.
Step 3 — Clone your first voice
- Open VoiceStudio and go to the Voice cloning workspace.
- Add a clean reference recording — 10 to 30 seconds of clear speech, no background noise, one speaker.
- Install the suggested model when prompted (one-time download).
- Type your text and hit generate. That's it — the audio is produced locally.
Tips for a better clone
- Quality in, quality out. A noisy reference recording is the #1 cause of robotic output.
- Record in a quiet room, 20–30 cm from the mic, at a steady pace.
- Longer references (30s+) improve prosody; 10 seconds is enough for a quick test.
- For audiobooks, generate chapter by chapter and keep the same reference clip throughout for a consistent voice.
Uninstall or upgrade
curl -fsSL https://voicestudio.sh/install | sh -s -- --uninstallThe installer preserves your settings, projects, and models, and upgrading is just re-running the install command. VoiceStudio is AGPL-3.0 licensed open source — this site is not affiliated with the project.
Next: how it compares to ElevenLabs →
Frequently asked questions
How long does VoiceStudio take to install?
About 10 minutes on a typical connection. The first model download can add several minutes depending on its size and your bandwidth.
Can I run VoiceStudio without a GPU?
Yes. CPU-only mode is fully usable, but generation is slower and the CPU PyTorch build needs about 5 GB of free disk.
Is VoiceStudio free?
Yes. It is open source under AGPL-3.0; you pay only for your hardware and electricity. Optional remote services are separate.
Does VoiceStudio work offline?
Local workflows work offline after the app and models are installed. Optional remote services require an internet connection.
Sources
- VoiceStudio README — install commands, hardware table, and feature list (checked October 20, 2026).
- VoiceStudio releases — download formats and current version.
- VoiceStudio license — AGPL-3.0.
Related guides
ElevenLabs alternatives
Seven open-source options compared by license, platform, and use case.
Voice cloning for audiobooks
Long-form workflow, cost comparison, and platform caveats.