Dictating text

What this does

Task menu ▸ Dictate

You speak, and the text is inserted continuously at the cursor position. Unlike transcription you do not wait until the end – you see what arrives as you talk.

The settings

• Model – which model does the speech recognition.
• Language – the language you are speaking.
• Pace – the setting that makes the difference.

Choosing the pace

Three levels, and they are all about the same trade-off: the shorter the segments, the sooner the text appears – and the less context the recognition has to interpret a word correctly.

• Faster (less context) – text appears promptly, more errors on similar-sounding words.
• Balanced (recommended) – the default. Start here.
• More accurate (longer waits) – the text appears with a delay, but more of it is right. Sensible for technical text and proper names.

The pace is a setting on this device. You decide it yourself and it is remembered.

How to dictate

1. Put the cursor where the text should appear.
2. Start dictation.
3. Speak. The display alternates between Listening… and Transcribing….
4. Stop dictation.

Speak in whole sentences and pause briefly at the end of each. Half-sentences with a fresh start in the middle arrive unusable.

Say the punctuation you need. Try a few sentences at the start to see how your model handles it.

What to do afterwards

Dictated text is raw material. It has the structure of speech: long sentences, repetition, filler words. No Filler Words and Improve in the Improve menu are the obvious next steps.