Personal dictionary

Adding words

Open Dashboard ▸ Dictionary and type the word as it should be spelled, then press add. Entries are stored immediately.

Good candidates:

• Colleagues' and clients' names, especially non-English spellings.
• Product and project names — yours and your customers'.
• Technical terms and command names: kubectl, WhisperKit, TestFlight.
• Anything you have corrected by hand more than twice.

There is no cap on entries, but a focused list works better than an exhaustive one — every entry is context the cleanup model has to weigh.

How entries are applied

The dictionary is not a find-and-replace pass. Your words are supplied to the cleanup model as explicit instructions — roughly “always use these exact spellings” followed by your list — so the model corrects a misheard variant in context rather than blindly substituting text. That is why an entry can fix a name even when Whisper produced something only loosely similar.

Entries may also carry aliases: known mishearings of the word, which are passed along as “may be misheard as…”. Aliases are shown in the dictionary list when present. The Know-Me interview seeds some entries automatically from the names and jargon you provide.

Duplicates and editing

Adding a word that already exists (ignoring case) merges into the existing entry rather than creating a second one. To remove an entry, use the delete control on its row.

Where it is stored

~/Library/Application Support/OpenVoiceFlow/dictionary.json, written with file protection on every change. It never leaves your Mac except as part of the cleanup prompt, and then only to the provider you chose. See Privacy architecture.

Seeding from your profile

Running the Know-Me interview adds the names and technical terms you list into the dictionary automatically, without overwriting entries you added yourself. It is the fastest way to populate a useful starting set.