VoiceStudio launches open-source local ElevenLabs alternative with 646-language voice AI

VoiceStudio has emerged as a fully open-source, locally-operated alternative to ElevenLabs, packaging voice cloning, voice design, video dubbing, dictation, transcription, and audiobook creation into a single Electron desktop application. The project defaults to the k2-fsa/OmniVoice engine but supports multiple backends including optional GGUF builds with Vulkan acceleration on AMD GPUs. It targets broad hardware compatibility, offering CUDA for NVIDIA GPUs, Metal for Apple Silicon, ROCm opt-in for AMD on Windows, and CPU-only operation with a minimal PyTorch build for systems without discrete graphics.

The project provides one-command installers for macOS and Linux via shell script, PowerShell for Windows, and distributes packaged builds as .dmg, .exe, .AppImage, and .deb. Version 0.5.3 marked the final Tauri release; the application has since migrated entirely to Electron with the Tauri shell and legacy UI entry points removed. The stack requires Node.js 22+, Bun, Rust/Cargo, and Python dependencies for source builds, with a managed runtime installation path available for end users.

Architecture, engines, and hardware acceleration

• Electron-only desktop stack: The project completed its migration from Tauri to Electron; version 0.5.3 was the final Tauri release and all legacy Tauri shell code has been removed, unifying the desktop and web UI on a single runtime.
• Multi-engine voice pipeline: Default inference uses k2-fsa/OmniVoice with swappable engines including CosyVoice GGUF, which can optionally leverage Vulkan on Windows and Linux when built with the community runtime; verified on AMD RX 6700 XT.
• Hardware abstraction layer: Supports NVIDIA CUDA on Windows and Linux, Apple Metal Performance Shaders (MPS), AMD ROCm as opt-in on Windows 11, and a CPU fallback that installs a reduced PyTorch build requiring approximately 5 GB disk space.
• Local API and agent integration: Exposes a local API and Model Context Protocol (MCP) endpoint for coding agents, with documented integration paths for Claude Code, Codex, and Cursor via install prompts and npx skills add debpalash/VoiceStudio.
• Build and distribution system: Source builds require Git, Node.js 22+, Bun, Rust/Cargo, and platform toolchains; releases are SHA-256 verified with an installer that preserves user data, projects, and models across updates.

Why it matters

VoiceStudio addresses a significant gap in the open-source audio AI stack by consolidating fragmented voice tools into a single local application with broad language coverage, reducing reliance on proprietary cloud APIs like ElevenLabs. Its AGPL-3.0 licensing, combined with explicit model-source provenance tracking and consent-based analytics, offers Linux distributions and privacy-conscious users a transparent alternative for content creation workflows. The project's investment in cross-platform hardware acceleration and agent-ready APIs positions it as a practical tool for developers building automated media pipelines without surrendering data to remote services.