Vois positions itself as a local alternative to cloud text-to-speech services. Instead of paying per character and re-paying every time you regenerate a line, you install Vois Studio, download the voice models you want, and generate as much audio as your computer can handle. The pitch is straightforward: a fixed monthly or annual subscription replaces metered credits, so previewing a narration twelve times costs the same as doing it once.
The app is built as a production environment, not just a TTS box. It includes a script editor that keeps chapters, speakers and takes organised, multi-speaker casting so each character can get a distinct voice, a pronunciation dictionary for names and unusual terms, and a multi-track timeline where narration, music and sound effects are balanced with EQ and loudness tools before export. Presets cover Spotify, Apple Podcasts, YouTube and audiobook stores such as Google Play Books and Kobo.
Beyond the core studio, Vois ships CLI access and skills for local coding agents like Claude Code and Codex, so an agent can draft chapters and then request generation and export steps that you approve. Voice tooling includes cloning from a short recording and a Voice Design feature that creates a new voice from a text description; on the Pro tier, cloned and designed voices are reported to speak all 30 supported languages while keeping their accent. The free tier is limited to non-commercial work with around ten generations per day, and the vendor notes that the Omni engine benefits from a compatible GPU and roughly 16 GB of memory.
