Voice Styles

What you say and what ends up on the page are not the same thing. On the Voice Styles page you decide how far apart the two are allowed to be. You choose there using examples rather than descriptions: next to every option is the same sentence, written the way that option would write it.

The baseline

The baseline applies everywhere you dictate. It has three levels:

Raw. The text is there instantly, exactly as recognised, filler words and all, with no tidying up.
Light. About one and a half seconds. Filler words go, punctuation and capitalisation are right, the sentences stay yours.
Strong. About three seconds. On top of that, sentences are pulled together and rearranged where that makes the text clearer.

Light is the default.

The baseline levels are named the same as the two recognition tiers, but they mean something different. The baseline decides how much a finished text is edited. The tier under Recognition on your machine decides how your voice becomes text in the first place.

On Raw nothing is sent off for editing. If you want raw as a rule but a single dictation edited anyway, use Ctrl + Shift + Space for that one.

The styles

Below the baseline you set a finer option per kind of application. All three styles are on in the shipped state.

Mail for Outlook, Thunderbird and Mail. Three options: Formal writes out contractions and keeps a formal business tone, Matter-of-fact writes full sentences with clean paragraphs, Relaxed keeps the sentences short and the tone direct. Matter-of-fact is the default. In all three, a dictated salutation stands alone on the first line, followed by a blank line. A salutation or a sign-off is never invented.
Chat for Slack, Teams, Discord, WhatsApp, Telegram and Signal. Three options: Matter-of-fact, Relaxed and Very relaxed, which writes everything in lowercase and uses little punctuation. Relaxed is the default. Extra paragraphs, salutations and sign-offs are never added here.
Editor for Visual Studio Code, Visual Studio, IntelliJ and Cursor. Technical terms, file names and code stay exactly as you spoke them, and nothing is reformatted.

In every other application the baseline applies and nothing else.

How Stenowise knows where you are dictating

From the name of the application you were in when you pressed the key. Window contents and window titles are never read.

The name leaves your machine only when a style actually applies to that application. Dictate into a program no style covers and the name does not travel either.

Switch off "Use the application context" in the settings under Data and privacy and no style applies any more, leaving the baseline everywhere.

When the editing does not come back

The editing has a time limit of nine seconds. If nothing comes back within it, the raw transcript is inserted so you are not left in front of an empty field.

The same happens if two attempts come back empty. A card at the edge of the screen then tells you this is the raw version. Your replacements from the dictionary are applied to it all the same, they do not hang on the editing.

If your weekly allowance ran out mid-dictation, you get the raw transcript for that one too. More on that under Account.

Switching afterwards

In the history you can switch any entry between the edited version and the raw one, back and forth. Both are stored for as long as the entry is there.