Voice Diary
Speak your day. We'll keep it.
In development · private beta on TestFlight · not yet on the App Store
Voice Diary is a journal you talk to. You tap a button, say what you want to remember, and the app writes it down — no typing, no blank page to fill in. Over a day those fragments pile up; in the evening they are woven into one connected text you can read back, edit or rewrite.
The point is what happens later. A week or a month on, the app can look across everything you said and tell you what it noticed — which themes kept returning, what shifted, what you never resolved. You can ask it questions about your own year and get an answer grounded in your own words.
The diary itself stays on the device. There is no account and no server-side copy of your entries — only the text being summarised at that moment is sent to the model, and the app can export everything you have written as a single file.
Inside the app
Real screens from the current build. Tap any of them to enlarge.



Screens are shown with sample content.
How the AI works here
What the model actually does in Voice Diary, screen by screen.
Speech to text
Every recording is sent to gemini-2.5-flash as inline audio and comes back as a clean transcript, with the spoken language detected automatically. A native on-device recogniser runs alongside it purely to show words as you speak; the text that is kept is always the model's.
The daily summary
At the end of a day the model receives the raw transcripts with their timestamps and writes one chronological text of up to 700 words. Reasoning tokens are switched off for this call so the whole budget goes into the summary itself.
The monthly synthesis
Once a month the daily summaries — plus the previous months for context — are turned into five sections: the arc of the month, recurring themes, what shifted, what carried over, and the questions left open.
A one-line insight
The month's synthesis and your own messages from reflection chats are condensed into a single sentence of at most 130 characters, shown when you open the app.
Reflection in conversation
Chat replies stream over SSE from gemini-2.5-flash. The model is given a budgeted context — up to 24 monthly syntheses, the last fourteen days, and today's raw notes — so it answers from your actual diary.
Reading a paper diary
Photograph the pages of an old notebook and the same model reads them, holds a conversation about what it found, and returns dated entries as structured JSON, ready to import into the calendar.
At a glance
- Platform
- iOS
- Languages
- 16
- Version
- 1.0.1
- Models
- gemini-2.5-flash
Voice Diary is not yet available for download. When it ships, the store link will appear here.