Use case P68 · Human
Parent builds a voice journal with an on-device speech model to record family memories without typing
- Family, travel and memories
- Advanced
- The Early Days, transformers.js, Supabase
- Source checked · blog post
- Vendor source
The short version
What they did
Cory Trimm built The Early Days for recording memories about his daughter by voice instead of typing. He says a speech-recognition model runs in the browser through transformers.js, producing searchable entries with dates and tags. The app also uses Supabase for accounts and storage, so on-device transcription does not establish that the entire journal stays offline. He calls it private and in active use, not a generally available app.
Result (as claimed)
“I built The Early Days - a voice-first journal for new parents.”
“Tap and talk, get a written, searchable record of the early weeks.”
“Currently in active use.”
Prompt
Adapted build request, not a prompt Cory published or a claim that he used a coding AI:
I want a simple voice journal for my family’s memories. I’m asking Claude Code to build a first version for [device and browser], using an on-device speech model through transformers.js. My chosen storage location is [where recordings and text should be kept].
1. Plan one record button, an editable transcript, a date and optional tags. Explain what runs in the browser and where each kind of data is stored.
2. Build the smallest version and test it with [a made-up memory].
3. Show how to find an entry by date or tag.Give me a short setup guide and a table of test results in plain words. Ask before adding uploads or changing the storage choice. Do not describe the whole journal as offline just because transcription runs on the device.
Do it yourself
These steps are my suggested way in, not Cory’s setup.
Level: Advanced to build your own app. The first version needs no code.
What you need
- A phone with a voice recorder or built-in dictation.
- A notes app or a document for the entries.
- For the build, a coding agent such as Claude Code, and a browser speech model through transformers.js.
Start simple
- Record a one-minute voice memo about something your child did today.
- Use your phone’s dictation or the recorder’s transcript to turn it into text.
- Paste it into a dated note. Add two or three tags, such as “first words” or “grandma”.
- Do this for a week and see if you keep it up.
Take it further
- Ask a coding agent to build a one-button web page that records, transcribes in the browser and saves a dated entry.
- Decide where entries are stored before building. Start with storage on your own device.
- Add search by tag and date, then an export to a plain file you can keep.
Check your work
- Read each transcript and fix names. Speech models get names wrong.
- Test with made-up notes first. Keep your child’s details off any cloud service until you know where they go.
- Back up the export somewhere you control.
Source checked. The link, author, date and quote were confirmed against the source. The result itself has not been reproduced. Prompt adapted for this library, inspired by the case. The source was published by a vendor of the tool. Business size: Individual. Results are quoted as the source claimed them and never improved. Spot an error or a broken link? Email charlie@cbarmorecpa.com.