Dictionary
Last updated
Add names, jargon, and technical terms to the OiPer dictionary so voice dictation spells them correctly, and see which transcription modes use it.
The dictionary is a list of words you often say that speech models tend to mishear — personal and product names, company names, technical terms, and acronyms. The list is passed to the speech model as a hint, which makes it far more likely to spell them the way you intended.
Adding words
On the Dictionary page, type a word and add it. Add the exact spelling you want in the
output, including capitalisation, for example PostgreSQL rather than postgresql.
Short multi-word phrases work too if that is how you say them.
Size limit
The whole dictionary is limited to 896 characters, counted as each word's length plus three characters of overhead. The page shows how much of the budget you have used, and stops you once it is full.
Because the budget is small, spend it on words the model actually gets wrong. Ordinary vocabulary does not need to be listed.
Which modes use it
The dictionary is applied in cloud and oiper modes.
Local Whisper transcription does not currently use it. If you dictate in local mode and a
name keeps coming out wrong, use a snippet to rewrite the common
misspelling instead, or move that profile to a larger Whisper model.
Snippets
Create OiPer snippets that turn short spoken shortcuts into longer text when you dictate, and learn how matching decides which snippet is applied.
History
Review past dictations in OiPer, play back the recorded audio, see the details of each run, and clear your local transcription history at any time.