What Aurora's speaker attribution actually scores
Measured word-level speaker error on four public-corpus suites, the stage of Aurora's own pipeline that damages it, and the downloadable per-case results behind every number.
Blog.
How Aurora works, what stays on your device, and what we are still improving.
How Aurora decides a live word is safe to show
Words appear as you speak, and the newest ones sit in a dimmed tail until two decodes agree and 0.6 seconds of audio has been captured past them. The rules, the constants, and what changes after you stop.
What happens to your audio after you press record
Aurora writes captured audio to one protected directory on your own device, twice, before any model reads it. The file path, the format, the crash-recovery sidecar, and every network call the app is allowed to make.
What is actually inside an Aurora diagnostic report
One switch decides whether a report carries your transcript, chat, recording and screenshot — and on a session report it starts on. Every tier, and the real members of the package.
Why Aurora does not send a bot to your meeting
Put your iPhone in the room and record through its microphone. What that buys, what it gives up, and which setups it cannot hear.
What on-device transcription costs on an iPhone
Offline processing uses your phone's storage, memory and battery. Here is the one throughput figure we have measured, exactly what Aurora downloads, and the numbers we do not have yet.
Build notes, when something changes.
A short email when a build ships or a measurement moves. You confirm once; we never add you to anything else.
AURORAYOUR WORDS. YOUR WORLD.
