Summary
Alpha Stack demonstrates how to set up free, locally-run voice cloning on a Windows PC using VoiceBox (version 0.1.12), an open-source application by developer Jamie Pine. The synthesis engine is Qwen 3 TTS at 1.7 billion parameters, running entirely on-device with no cloud dependency, subscription fees, or per-character cost — positioning it as a practical alternative to paid services like ElevenLabs.
The tutorial covers the complete installation flow: downloading VoiceBox from Jamie Pine’s GitHub, handling the unsigned publisher warning on Windows, and waiting for the initial 5GB Qwen 3 model download on first launch. Voice profile creation requires uploading a reference audio clip alongside its transcript; the system generates a cloneable voice profile in under a minute. The creator demonstrates cloning both their own voice and a Trump audio recording, observing that punctuation — commas, question marks, ellipses — meaningfully affects prosody and intonation in the output.
On performance, single-word generation takes roughly 20 seconds and a full sentence around one minute on the creator’s hardware — slower than cloud alternatives but acceptable for offline use cases. Audio quality is rated as solid for a free local solution. The key tradeoffs are generation latency versus zero cost and complete privacy, since no audio data leaves the user’s machine. The video includes a downloadable cheat sheet in the description covering all required links and setup steps.
📺 Source: Alpha Stack · Published February 19, 2026
🏷️ Format: Tutorial Demo
