"Send to Cassettoni."
Auris hears the word, doesn't know it. Best guess lands in the app.
Formality
Casual keeps contractions and a friend-tone. Professional uses full grammar and a measured tone.
How it works
How Auris works.
The basic loop in three steps.
Everything else is how you shape what comes out.
The loop
Hold. Speak. Release.
Hold the hotkey, talk, release.
The full loop runs in under two seconds.
The full loop runs in under two seconds.
01 · You
Hold, talk, release
⌃fn Mac
⌃Win Windows
⌃Win Windows
02 · Auris
Whisper / Groq transcribes
~200ms after release
03 · Auris
LLM polishes
~500-900ms
04 · Result
Clean text in active window
~1.5s total
Local-only path: Whisper-local + Ollama means audio and transcript both stay on your machine.
Sliders
Three sliders that change the output.
Auris ships with three global tone sliders. Move them.
The next dictation reflects the change.
The next dictation reflects the change.
live demo
same dictation, drag to repolish
You should probably try the new Italian place on Main, food is really good and prices are reasonable.
Dictionary
Fix it once.
When you edit a polished output, either in Auris's history view or by fixing a typo within five minutes of dictating, Auris notices the diff.
"Send to Castoni."
You fix the typo in whichever app it landed in. Auris works in every app where you can type.
Castoni → dictionary
The correction goes into your dictionary automatically. Next time you say it, it lands right.
The dictionary catches matches before the polish step ever sees them. You stop fixing the same name, acronym, or technical term over and over.
Snippets
Spoken triggers, written shortcuts.
Snippets fire on phrases you say. Say "today's date," get the current date. Say "meeting template," get a notes skeleton with the date already filled in.
Snippet variablesauris/snippets
{date}today's date→May 18, 2026
{signature}your name and title from Settings→Jane Doe · VP Design
{time}current time→3:42 pm
{clipboard}paste from your clipboard→contents of ⌘V
Build your own snippets in the Dict tab. Edit them whenever.
They live on your machine.
Sensitive
Sensitive content stays out of your history.
When you dictate something that matches a privacy pattern, the polished text lands in the app you're typing in like any other dictation. Auris skips writing it to history.jsonl. No local copy.
Default patterns
email
phone
SSN
credit card
Add your own with regex for company-specific things: employee IDs, internal project codes, internal URLs. See the full pattern list on the privacy page.
Note: Patterns match locally after the transcript returns. Cloud STT or cloud LLM polish sees the unredacted content first; only your local history gets filtered. For fully off-network protection, run both speech-to-text and polish locally.
Setup
Two minutes to start.
First time you open Auris, a setup walks you through four steps.
None of them take long.
None of them take long.
Microphone access
Click "Allow" when the system prompt appears. On Windows, you might also need to grant access in Settings → Privacy → Microphone.
Pick your polish path
Choose where Auris runs the polish step: OpenAI, Anthropic, or Groq with your API key, or a local model for full privacy.
Confirm hotkeys
Ctrl + Fn on Mac, Ctrl + Win on Windows by default. Change them if you want.
Start talking
Hold the key. Say something. Watch your words land in the text field.
See it run on your machine.
Auris is in beta. One email when it ships, no newsletter.
The setup is two minutes; the first dictation is one keystroke away.
Mac & Windows · Lifetime updates