An interview voice is not a brand asset — it is the thing your users listen to for five minutes. Browse the full catalogue, filter it, and hear any voice on a real KKOS line before you commit.
An editorial shortlist from the catalogue below — each one leads a different kind of conversation. Click to hear it. These are real catalogue voices, not separate recordings.
Metadata comes straight from the voice provider. Auditioning synthesises a short KKOS line through the same endpoint the product uses.
Both panels are computed in your browser from the same catalogue array that fills the grid above — a second reading of the same data, not a second source.
Six largest groups. Everything below the sixth sits in the single digits.
Share of voices that carry each field at all. The gaps are absent in the catalogue data itself — they are not filtered out by this page.
Source: the voice catalogue embedded in this page, captured from the voice endpoint. Both panels are recomputed on every load, so they follow catalogue updates instead of going stale.
Expressive, emotionally aware speech synthesis. The default for every interview voice above.
—Free, fast, always-available fallback. Used automatically if the primary engine is unreachable, so a session never dies on voice.
Always availableSelf-hostable conversational engine with style controls — for teams that must keep inference on their own infrastructure.
—Roadmap: natural-language voice design (describe a voice, get a voice) and reference-clip voice cloning are not shipped yet. This page only lists what you can actually play today.