Keep a speech dictionary

Speech recognition writes down what it hears. Names, places and project jargon it has never met come out misspelled — Project Waddle as "Project Wattle". The speech dictionary is the list of words Lotti should get right, kept once for the whole app.

All screenshots

Each term shows where it applies. Search also finds a term by a spelling it was misheard as.

ScreenshotMobile view

Add a term

1. Open Settings → Definitions → Speech dictionary.
2. Select Add term, or type the word into the search field and choose Add "…" when nothing matches.
3. Enter the term exactly as it should be spelled.

You can also select a word in the text of any entry and choose Add to Dictionary. It is added for that entry's category.

Limit a term to categories

All screenshots

With no category chosen, a term applies everywhere.

ScreenshotMobile view

Use Choose categories to apply a term only to recordings in those categories, so niche vocabulary stays out of unrelated transcripts. With none chosen, the term applies to every category.

Misheard as

List the spellings a term tends to come out as, separated by semicolons. Lotti adds the ones it corrects, too. They are hints, never automatic replacements: a misheard spelling can also be a real word the speaker meant, and the surrounding text decides.

How transcripts are corrected

When a recording in a task is transcribed by a speech-to-text engine such as Whisper, the step that writes its one-liner and summary also corrects the dictionary terms the transcript misheard. The recording's text appears once, already corrected, and stays editable. A multimodal model that listens to the audio itself reads the dictionary in its own prompt instead.

The terms your categories held before were carried over automatically, each limited to the categories that held it.