Voicebox setup: voice cloning with Qwen3-TTS

This tutorial explains how to install Voicebox on Windows and create a voice profile for text-to-speech with Qwen3-TTS. The presenter demonstrates local AI speech generation using a recorded sample, then compares the 1.7B and 0.6B models.

The setup covers checking PC specifications, downloading the appropriate installer and waiting for the initial environment preparation. Users can record through a microphone or import an audio file. The presenter enters reference text, selects Korean and leaves the Personality field blank to have a completed script read as written. They explain that the sample's emotion and intonation influence the generated voice.

The presenter describes 1.7B as the quality-focused option and 0.6B as suited to lower specifications and faster generation, recommending at least 6GB of GPU memory for 1.7B. Their CPU and GPU comparison reports a speed difference without an apparent quality difference. GPU setup involves downloading CUDA through settings, restarting and checking recognition.

For clipped sentence endings, the presenter suggests adding an extra word such as 'end'. The tutorial also covers basic audio edits and ends with a separate browser workflow for making text-motion videos. Claims about commercial use and the absence of generation charges come from the presenter.