← All repos

voicebox

The open-source AI voice studio. Clone, dictate, create.

GitHubvoicebox.sh
aicudamlxqwen3-ttsqwen3-tts-uivoice-aivoice-clonewhisper
Browse cluster: SQL databases and distributed systems →
50,591stars
638commits
77contributors
9languages

Tech stack & purpose

Voicebox is a local-first AI voice studio that provides voice cloning, speech generation, and audio transcription capabilities. The project encompasses a FastAPI backend built in Python that handles TTS and STT inference through multiple engines including MLX and PyTorch with CUDA acceleration, a Tauri desktop application for the user interface, a Next.js documentation site using Fumadocs, and a Next.js landing page. The backend detects the best inference backend at startup (MLX on Apple Silicon, PyTorch with CUDA on NVIDIA GPUs, with fallbacks for other platforms), implements a Model Context Protocol server allowing MCP-aware agents to access voice cloning and transcription tools, and organizes its API into 90 endpoints spanning profiles, generation, history, transcription, stories, effects, and model management. The repository is authored by Jamie Pine.

Languages

TypeScript
53.4%
Python
35.3%
Rust
8.7%
Shell
0.8%
Just
0.7%
JavaScript
0.4%
CSS
0.4%
Dockerfile
0.2%
HTML
0.1%

Contributors (top 30 of 77)