Voicebox: Italian voice cloning and dialogue editing

This tutorial covers VoiceBox installation on Windows, voice cloning and speech generation. The presenter describes the software as open source, free to use without subscriptions, and able to work offline on Windows, Linux and Mac. He positions it as an ElevenLabs alternative, but does not run a side-by-side comparison.

The voice profile workflow uses an uploaded audio file or a microphone recording. The presenter recommends about 30 seconds of clear speech without background noise, selects Italian, and transcribes the sample before creating the profile. Transcription is required to complete the demonstrated workflow. English preset voices are also available. He then generates speech with his cloned voice and calls the result faithful; that is his assessment of the sample. The first model invocation needs time to load, and he reports roughly a minute for his initial generation.

A second audio file supplies another voice profile for the dialogue editor. The demonstration covers assigning voices to lines, moving and overlapping tracks, and trimming clips. Vocal effects include robotic and deeper speech. The model panel shows downloaded models, with one described as occupying 4 GB. The presenter discusses a fidelity and processing-time tradeoff without providing a benchmark. He also explains that MCP can connect VoiceBox with Claude Code or Codex to send text and automate voiceover creation, though he does not demonstrate that automation.