Text replacements

Available onWebmacOSWindows

A text replacement maps one or more spoken forms to the text that should appear in the transcript. Speech models are consistent in their mistakes: if yours always hears your registrar’s name as “Dr. Cross” instead of “Dr. Krauss”, or spells a local place name wrong, fix it once with a replacement instead of editing every transcript.

How it works

Each replacement has:

• Source terms: one or more spoken variants to match (“doctor cross”, “dr cross”).
• Target text: what appears instead (“Dr. Krauss”).

Replacements are applied to the transcript automatically as it is produced. They can be enabled or disabled individually.

Good uses

• Names the model consistently mishears (colleagues, wards, clinics, local place names).
• Local abbreviations you want expanded consistently.
• Formatting you always want (“b.d.” vs “BD”).

Medication names have their own net

You can add replacements for drug names, but the desktop apps already run a medication safety net that matches garbled drug names against a medicines list. Reserve replacements for the words it does not cover.