Two modes. First, a listen button on every reply so every answer can be played back aloud. Second, a live voice call that streams audio both ways — the buyer asks, the AI answers, in real time, in their language.
Every AI reply shows a Listen + Copy control. Buyers tap to hear the answer read aloud in their chosen voice — great for buyers on the go or with accessibility needs.
The buyer taps the mic, the widget opens a realtime WebRTC stream to an AI voice model, and the two of them talk — the AI answers aloud, grounded in your inventory + documents.
Arabic-speaking buyers often prefer speaking over typing on mobile. Voice lowers the barrier to entry.
Walking through a showroom, in the car, at a site visit — voice is the interface when screens can't be.
Visually-impaired buyers, or buyers who read slowly, get the same answer in under a second.
Voice streaming is opt-in at both the platform level (env flag) and the tenant level (settings toggle). If the platform disables it, the settings section is locked out — buyers never see a broken button.
Voice sessions mint short-lived ephemeral keys on the server — no permanent credentials are ever exposed to the browser. WebRTC traffic is encrypted end-to-end. Per-tenant guardrails apply to voice replies exactly like text replies — no fabricated prices, no off-brand claims.
Start your workspace free today — voice streaming rolls out to every tenant as soon as it ships. No waitlist.