Problem

Medical and government documents are unreadable with a screen reader, because screen readers read characters, not structure.

Approach

Hear → Speak → Confirm → Complete → Download. Reads documents aloud by structure and fills forms by voice with ElevenLabs speech and Scribe STT; Linkup is used only for trusted terminology. Safety rules are baked in: every spoken value is read back and confirmed before a field is filled, low-confidence values are marked Needs review and never guessed, signature fields are never completed, nothing is stored.

Results

Falls back to the browser speech engine with no API keys, so it works anywhere.