niomi.ai proposes conversational avatars as an optional interface layer to make interactions warmer and more approachable. Users could choose visual style, voice, pace and tone—options that matter for accessibility, cultural fit and emotional comfort. The interface would continuously and transparently label the system as AI, and the avatar’s behaviour would be constrained by the same deterministic safety engine that governs text interaction: no independent diagnosis, prescribing, or unilateral emergency decisions.
Technically, the avatar layer overlays speech-to-text, text-to-speech and visual rendering onto the conversation interface while logging all actions in the audit service. Safety-critical outputs are routed through the clinical rules engine for verification before any escalation. When a high-risk signal is detected, the avatar transitions the user to a clinician-mediated pathway with clear steps and provenance — preserving empathy without blurring authority or accountability.
