Skip to content
Now with hardware-aware GPU acceleration on supported Windows PCs. Results vary by hardware and model.
EYRO

Privacy

Your voice and transcripts do not need to leave the computer for normal dictation

Local processing supports privacy-conscious workflows. It does not by itself make a customer’s workflow compliant with every regulation. Configuration, contracts, and process still matter.

Download EYROOpen the trust center

Path

How data moves

Microphone → local buffer → local Whisper → active field. Temporary recordings are deleted after attempts. Local history and deletion controls stay visible.

  1. 01

    Microphone

    On device

  2. 02

    Local buffer

    Temporary

  3. 03

    Local Whisper

    Local model

  4. 04

    Active field

    Inserted here

Inventory

Data inventory

What EYRO touches, where it lives by default, how long it stays, and whether it leaves the device for normal offline dictation.

  • Microphone audio

    Purpose
    Capture speech for local transcription
    Default location
    Temporary local buffer / file
    Retention
    Deleted after transcription attempt
    Leaves device?
    NoNo for normal offline dictation

  • Transcript text

    Purpose
    Insert into the active app; optional local history
    Default location
    Active text field; optional local history store
    Retention
    User-controlled history; deletable
    Leaves device?
    NoNo for normal offline dictation

  • Personal dictionary

    Purpose
    Improve names and specialist terms
    Default location
    Local settings store
    Retention
    Until the user edits or deletes entries
    Leaves device?
    NoNo by default

  • Settings & hotkeys

    Purpose
    Model choice, shortcuts, language variant
    Default location
    Local preferences
    Retention
    Until changed or uninstalled
    Leaves device?
    NoNo

  • Diagnostics / telemetry

    Purpose
    Reliability metrics without dictated content
    Default location
    Local first; optional outbound if enabled
    Retention
    Minimised; retention limits disclosed when enabled
    Leaves device?
    ConditionalOnly if you opt in or as separately disclosed

  • Updates & model downloads

    Purpose
    Keep the product current and install models
    Default location
    Downloaded to device
    Retention
    Installer and model files on disk
    Leaves device?
    ConditionalRequires connectivity for download/check

  • Purchase & licence metadata

    Purpose
    Payment, tax handling, licence delivery, activation, cancellation, and refunds
    Default location
    Paddle and the CAWOSH Cloudflare licence service
    Retention
    Minimised for licensing, fraud, support, accounting, and legal obligations
    Leaves device?
    YesYes when purchasing, activating, checking, cancelling, or requesting a refund

  • Optional online intelligence

    Purpose
    Grammar, paraphrase, summary, hosted models
    Default location
    Only when explicitly enabled
    Retention
    Per feature terms shown before use
    Leaves device?
    YesYes, only with explicit consent

May connect

Connectivity that may exist

  • One-time or update-time model downloads
  • Update checks and signed installer delivery
  • Paddle checkout, tax handling, licence delivery, activation checks, cancellations, and refunds
  • Optional telemetry or crash reports only when disclosed/enabled
  • Optional Advanced Intelligence: explicit consent, pricing, and terms

Not collected

What we do not collect by default

  • Raw audio
  • Transcript text
  • Dictionary entries
  • Active-window titles, clipboard, contacts, or field contents

Privacy

Questions about local processing

For normal offline dictation, audio is processed by the local engine. Temporary files are deleted after attempts. Exceptions are listed above and never silent cloud fallbacks for core dictation.