Recognition on your machine

By default Stenowise recognises through the server. You can have the recognition run on your own machine instead. Then the recording does not leave your device.

Where you switch

In the settings under Data and privacy, at "Recognition". The same switch sits in the menu of the icon in the taskbar. You are asked about it once on the first run.

The two tiers

Before your machine can recognise anything, one tier has to be downloaded. There are two:

• Light. Runs on any machine, including one with no graphics card, and answers immediately.
• Strong. The best result your machine can produce on its own. It needs a fast graphics card.

Strong is only offered if your machine has a graphics card up to it. Without one the tier is not in the list at all, not greyed out: it would recognise more slowly than you speak, and a dictation that arrives a minute after it was spoken is not a slower dictation, it is no dictation. Stenowise checks your machine before it shows you the choice.

The tier is downloaded once and sits on your disk afterwards. You can switch tiers at any time, and delete one too.

What changes

• The recording does not leave your machine. It becomes text here and is dropped afterwards.
• The recognised text still goes to the server for the cleanup, as long as the cleanup is on. Only with the cleanup off does nothing at all leave your machine. The four cases are under Privacy.
• You still have to be signed in. Signing in has nothing to do with where recognition happens.
• Words are still counted. The app reports a number and a length afterwards, never a word of the dictation. How the allowance works is under Account.
• After five minutes without a dictation the memory in use is released. Once you are back at the machine it loads again in the background, so your next dictation does not have to wait for it.

When this is worth it

When a recording must not leave your device. That is the reason the switch exists.

And when you work without a connection: with the cleanup off you can keep dictating offline, as long as this device is signed in.

For everything else, recognition through Stenowise stays the default and the faster way. It asks nothing of your machine and waits for no download.

When something does not start

With no tier downloaded, recognition stays with Stenowise and a card points you at the choice. So nothing happens quietly.

If Stenowise reports that recognition on this device did not start, pick the tier again under Data and privacy or switch back to Stenowise. Either is one click, and you can carry on dictating right away.

The full list

In the settings under Advanced you will find every entry recognition on this device can load. Day to day you do not need that list: there you choose between Light and Strong, and that is the whole decision.