VoiceStudio review: voice cloning vs ElevenLabs

VoiceStudio is an open source desktop app for voice cloning, speech generation and transcription. The presenter demonstrates a clone made from roughly 15 seconds of recorded speech, then tests expressive annotations. He also covers voice design through text prompts, a saved voice gallery and audiobook projects with multiple characters.

The dubbing tests give mixed results. A short Spanish sample sounds clear to the presenter, but a podcast attempt takes about half an hour, assigns his voice to his co-host and fails to translate the audio as expected. He also reports a failed 10-minute video with unhelpful logs. These results limit his comparison with ElevenLabs, particularly for recordings with multiple speakers.

For people exploring local AI, he reports roughly 15 to 17 GB of RAM use during demanding tasks and recommends 24 GB as a practical starting point. His later suggestion that a 16 GB Mac Mini might work remains speculation. Settings include model selection and remote inference on another machine, plus an API, MCP access and a dictation skill for coding workflows.

The presenter cannot get the system-wide dictation shortcut working, which prevents VoiceStudio from replacing his usual dictation app. Although he describes audio processing as on-device, he also reports telemetry in the free version, so his opening claim that nothing leaves the machine needs that qualification.