The open-source AI voice studio. Clone, dictate, create — seven TTS engines and Whisper transcription running entirely on your device. No API keys, no subscriptions, no cloud.
Built by Jamie PinemacOS · macOS (Apple Silicon recommended) or any system with PyTorch / CUDAVoicebox is open source and runs inference locally — Vendo just helps you find and install great OSS tools. No credits, no API keys, no cloud.
Inference happens locally — no cloud, no backend, no latency.
Voice, models, and files never leave your machine.
No credits, no subscriptions, no API keys — just install and go.
Qwen3-TTS, Qwen CustomVoice, LuxTTS, Chatterbox Multilingual, Chatterbox Turbo, HumeAI TADA, and Kokoro — all bundled, all running locally.
Record a short sample or upload a clip and generate speech in that voice. Your audio never leaves your device.
Drop audio files or dictate live — transcription happens locally with OpenAI Whisper (standard or Turbo) via PyTorch or MLX.
Hold a hotkey, speak, and have the transcribed text typed into any app. Works anywhere macOS accepts keyboard input.
Models and voice data stay on your machine. No API calls, no cloud uploads, no telemetry — 100% offline once installed.
Ships MLX-optimized weights for M-series Macs and CUDA builds for NVIDIA GPUs. CPU fallback works but is slower for larger models.
Voicebox runs entirely on your machine — no servers to pay for, no credits to top up, no rate limits to hit.
Pay as you go — no subscriptions, no per-seat fees. Every model we bill through is listed in our rate card.