Before you begin
VoiceStudio installed on your computer, your own voice cloned from a clean recording, your first voiceover generated offline, and Claude (or Cursor/Codex) connected to VoiceStudio over MCP so your agent can make voiceovers for you.
Tools used
- VoiceStudio (free, open source, AGPL-3.0)
Claude Code, Cursor, Codex or any Streamable HTTP MCP client (optional)
Requirements
- Apple Silicon Mac with macOS 13.3 or newer, or a supported Windows/Linux PC. Intel Mac supports the app UI with a remote backend, not local generation.
- For macOS, plan roughly 10 GB free for the app, runtime and models; allow more for additional models and your audio projects.
- A quiet room and a phone or mic to record 30 to 60 seconds of your own voice
- Optional: Claude Code, Claude Desktop, Cursor or Codex if you want to use the MCP server
On this page · 9 steps
Your progress
0 of 9 steps
The walkthrough
Step 01 / 09
What VoiceStudio is
VoiceStudio brings voice cloning, voice design, transcription and video dubbing into a local desktop workspace. Start with your own voice and a short test script rather than a full production video. The project advertises 646 languages, but that is not a guarantee of equal quality or support across every engine and task. Local workflows use your hardware; remote services are optional, and analytics requires consent. Check those settings before using private recordings. Official overview: https://github.com/debpalash/VoiceStudio
Step 02 / 09
Check your computer first
VoiceStudio runs best with hardware acceleration: Apple Silicon Mac (M1 or newer): Metal acceleration. Windows or Linux with an NVIDIA GPU: CUDA acceleration. No dedicated GPU: fully usable on the CPU, just slower (the CPU setup needs about 5 GB of free disk). Intel Mac: the app UI only, connected to a remote backend. The big one is disk space. On first launch the app downloads its local runtime (Python and AI libraries, several GB), and on our Mac it refused to continue until at least 9 GB was free. Voice models download separately when you choose them (the default OmniVoice model is about 2.3 GB). For macOS, the current installation guide recommends about 10 GB free and macOS 13.3 or newer. Storage needs vary by engine. Leave headroom instead of treating 9 GB as a universal minimum. Official Mac requirements: https://github.com/debpalash/VoiceStudio/blob/main/docs/install/macos.md
Continue from here