Voicebox clones a voice from 3 seconds of audio locally on PC

The open-source Voicebox tool from Spacedrive clones a voice from 3–30 seconds of audio and runs locally on Windows, Linux and macOS. It uses Whisper and Qwen3-TTS models, supports 23 languages and requires no account or identity check.

• Voice cloning works from 3 seconds of audio, 10–30 seconds recommended
• Minimum specs: 8GB RAM, 5GB storage and a multicore CPU
• Supports 23 languages and five speech engines including Qwen3-TTS
• Since v0.4.5 cached models generate speech without internet