hush·hushhushhush

Why type
when you can speak?

The privacy-first voice input that turns natural speech into organized, ready-to-use text, while your voice and data stay on your device.

Get hush·hushNo account. No sign-up. No subscription.

Polishes up and organizes what you said.

Removes filler words. Handles self-corrections. Adds punctuation. Pulls bullet lists and paragraph breaks out of long thoughts. You get text you can paste, not a literal transcript.

What you said

We have three things to do, um first we need to grab the wine, then pick up the cake, no actually cake first then the wine, and uh don't forget to text everyone the address

What you get

We have three things to do:

  1. Pick up the cake.
  2. Grab the wine.
  3. Text everyone the address.

Smart Fix learns the words you correct.

Correct a misheard word once and Smart Fix remembers. Next time you say it, the right form lands automatically. No training run, no upload, no fine-tuning. Just a local dictionary that gets you.

01You dictate. The speech model heard ‘la boo boo’. Close, but not what you meant.
Journal
Picking up a la boo boo at Pop Mart this weekend.
hush·hush pasting panel showing the first dictation 'Picking up a la boo boo at Pop Mart this weekend.' — the speech model has never heard of Labubu so it fell back to a phonetic guess. No Smart Fix corrections yet.
02You correct it once. Smart Fix remembers.
Journal
Picking up a la boo booLabubu at Pop Mart this weekend.
hush·hush auto-learn toast: sparkles icon, label 'Auto-learned a word', strikethrough 'la boo boo' arrow 'Labubu', Undo button on the right.
03Next dictation. ‘Labubu’ lands automatically. Smart Fix applies the learned form before it ever hits your doc.
Journal
Picking up a Labubu at Pop Mart this weekend.Labubu collection is getting out of hand.
hush·hush pasting panel showing an auto-applied Smart Fix correction la boo boo → Labubu, with the Apple Intelligence gradient pill on Labubu and the Smart Fix footer beneath.

Matches the line you were already writing.

The app reads what's just before your cursor, then uses that to keep tone, pronouns, and proper names consistent in what it pastes. The dictation feels like the same hand kept typing.

Reply to Mei

Hey Mei,

Thanks for the ramen spot rec last week.

↑ The app read a slice of text before your cursor.

Just went last night, broth was unreal, definitely going back this weekend.

hush·hush pasting panel showing a context-aware sentence composed in the casual tone of the email above (reply to a friend about a ramen rec).

Multilingual, even when you mix.

Pick a language and dictate. Mix languages mid-sentence and the transcript stays accurate.

Chat
Taylor Swift的Vegas演唱会门票终于抢到了,VIP package贵到离谱,但绝对值。
hush·hush pasting panel showing a Chinese + English code-switched sentence — 'Taylor Swift的Vegas演唱会门票终于抢到了,VIP package贵到离谱,但绝对值。' — transcribed accurately across the language boundaries.
EnglishHeading out for a hike this weekend.
中文今晚去看新出的那部电影。
日本語週末は京都に行ってくる
hush·hush

Privacy by design, not by promise.

The speech and organizing models both run on your device. Your audio is processed locally, then discarded. Nothing is uploaded, recorded, or seen by us. No cloud, no accounts, no analytics, no telemetry.

Just speak.
The rest stays with you.

Get hush·hushNo account. No sign-up. No subscription.