Skip to content

Crayonic Voice Assistant

Status: In development. Details and timelines are preliminary. Everything on this page is planned unless stated otherwise.

Where this fits

Crayonic Voice Assistant builds on the Crayonic Secure Audio Service (Option 4 in the Solution Architecture), which already provides encrypted recording and transcription, and on Crayonic Comms for delivery.

Overview

Crayonic Voice Assistant brings AI-assisted messaging to the Crayonic wearable ecosystem on top of the encrypted recording and transcription that the Crayonic Secure Audio Service provides today. Its two functions are AI post-processing of transcripts (summaries, structured notes) and AI-powered message routing that turns spoken input into the right action on the right channel, bound to the user's hardware identity.

Planned Capabilities

Transcription and Post-processing

  • Recording and transcription — provided by the Crayonic Secure Audio Service: the Crayonic Badge records and encrypts on-device, and transcription runs through the speech-to-text provider you configure (local Whisper, Praxy Santé, or another provider through the adapter interface). Available today.
  • LLM post-processing — summaries, structured clinical notes and data extraction generated from a transcript by a configurable LLM provider. The Secure Audio Service already reserves the provider configuration for this; the processing stage itself is planned.
  • Identity-bound records — transcripts tied to the recording badge's identity, producing tamper-evident conversation records suitable for compliance, audit and handover
  • Selectable providers — organizations choose the STT and LLM providers (cloud or on-premises) that match their data-residency and privacy requirements

AI-Powered Message Routing

  • Voice-to-action — spoken input is interpreted by the LLM to route the correct message to the correct channel (individual, group, role, zone)
  • Integration with Crayonic Comms — routed messages are delivered as identity-bound, signed communications
  • Zone and role awareness — combined with Crayonic RTLS and Crayonic Device Manager roles, the assistant can target "everyone in Zone A", "the on-call nurse" or "security lead", from a single spoken request
  • Safety actions — critical phrases can trigger a Panic Button event or an emergency broadcast without a separate UI step

Security and Privacy

  • Voice capture requires user consent and is bound to the Crayonic device holder; recordings are end-to-end encrypted as described in the Crayonic Secure Audio Service
  • Transcripts and routing decisions carry an audit trail in Crayonic Device Manager
  • Organizations control which STT and LLM providers are used and can keep processing on-premises where policy requires it
  • No voice data is retained beyond policy-defined retention windows

Integration Points

Status and Early Access

Under active development. Contact Crayonic if you want to explore transcription and AI-powered routing workflows bound to a hardware identity.