Voxlet is a native Mac app that helps you create and train a voice model from your own recordings, then turn typed text into speech that sounds like you.

Start with a quick mic check and read short sentences from a built-in script. Voxlet checks each take for noise, clipping, and dropouts, then helps you choose the voice that sounds most like you. Generate speech from a clean reference recording, or fine-tune your own model and compare the results side by side.

Everything runs locally. Your recordings, voice models, and generated audio stay on your Mac.

Powered by Qwen3-TTS (Apache-2.0) on Apple’s MLX, with LongCat-AudioDiT (MIT) as an alternative engine.

Free beta for Apple Silicon Macs running macOS 14 or later. Use your own voice, or another person’s voice with documented consent.