Fix names that voice typing keeps getting wrong
When a voice model keeps mishearing a name or a term, fix it once in Loquela's Voice Editor and later dictations write it your way. Everything stays on your phone.
Voice models learn from a lot of speech, but your colleague's name, your street or the product you work on may not be part of it. Loquela has three ways to fix that, and they work together. What you add stays on your phone.
Fix a word in the Voice Editor
The Voice Editor shows the words you just dictated, so you can fix one without moving the cursor.
- Open it: hold the microphone key. To have it open whenever you dictate, turn on Open automatically in Settings, Voice & Models.
- Pick another choice: tap the wrong word. The Voice Editor shows other words the model considered. Tap the right one.
- Say it again or type it: if none of the choices is right, tap Say again and repeat the word, or tap Type it and type it.
Write it your way every time
After you fix a name or an unusual word, Loquela offers Always write X as Y, with the wrong word and your word in place of X and Y. Tap it, and later dictations write your word instead.
Loquela does not offer this between two everyday words, such as "their" and "there", because that rule would also change sentences where the first word was right.
You can add these yourself too. In Settings, open Typing & Suggestions, then Personal dictionary. Under Voice replacements, enter what voice writes and what to write instead, then tap Add. Remove one there when you no longer need it.
Help the model hear it
With a Parakeet voice model, Loquela can also listen for your own words while you speak, so the model is more likely to get them right the first time.
- Add the word to your Personal dictionary in Settings, Typing & Suggestions.
- In Settings, Voice & Models, open your Parakeet model's Model Settings and turn on Listen for your custom words.
This is free. Transcription is a little slower while it is on.
With Full
With Full, Learn repeated fixes in Settings, Voice & Models puts first a choice you have picked three times for the same word. Full also opens every voice model in the catalog. A model that reads the whole sentence may handle names and sound-alike words better, so try a few with the names you say most.
In incognito mode and in private fields, Loquela does not offer or save fixes.