> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mauvely.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Dictating

> The global shortcut, the dictation overlay, and the three ways Sotto can put text into another application.

## The shortcut

**Meta+Alt+D** by default. Change or disable it in **Settings → Shortcut**; it also appears in your
desktop's own shortcut settings.

Two modes:

| Mode                 | Behaviour                           |
| -------------------- | ----------------------------------- |
| **Toggle** (default) | Press to start, press again to stop |
| **Hold to talk**     | Hold the keys down, release to stop |

<Note>
  Sotto uses the desktop's GlobalShortcuts portal, which Plasma 5.25 and later provide. If your
  desktop does not have it, disable the shortcut in Settings and bind `sotto --toggle` to a key in
  your desktop's own shortcut settings instead — it works identically.
</Note>

## The overlay

A small dark pill at the bottom of your **active** monitor:

* A live level meter, so you can see it is hearing you
* The live transcript, which updates roughly once a second while you talk
* A `LOCAL` badge — a standing reminder that nothing is being sent anywhere

The overlay **never takes focus** and never gets tiled by a tiling compositor. It is drawn as a
layer surface rather than a window, which is what makes both true by construction.

By default it appears on whichever monitor is active, re-evaluated every time it is shown. Pick a
fixed monitor in **Settings → Appearance** if you would rather it always be in one place.

<Tip>
  Animations can be turned off entirely in **Settings → Appearance** — worth doing if you find
  movement distracting, or on a machine where every frame matters.
</Tip>

## What happens while you speak

Sotto does not wait until you stop:

* A **pause of about 0.7 seconds** ends the current utterance and commits it.
* While you are still speaking, the in-progress utterance is re-transcribed roughly every 1.1
  seconds so the live transcript keeps up.
* Very short bursts are discarded as noise rather than transcribed as a stray word.
* Each committed utterance is fed forward as context, so names and terminology stay consistent
  across a long dictation.

The practical effect: partial text appears within about a second, an utterance settles about
0.7 seconds after you stop talking, and the whole thing is finished and inserted a fraction of a
second after you press the shortcut again.

## Getting text into an application

Wayland has no universal "type this" facility, so Sotto picks the best strategy available.
Configure it in **Settings → Output**.

| Mode                              | What happens                                                            | Needs                                                                        |
| --------------------------------- | ----------------------------------------------------------------------- | ---------------------------------------------------------------------------- |
| **Clipboard + paste** *(default)* | Sets the clipboard, synthesises `Ctrl` `V`, restores your old clipboard | `wl-clipboard`, plus either `ydotool` or a one-time RemoteDesktop permission |
| **Type it**                       | Types the text key by key                                               | `ydotool`                                                                    |
| **Clipboard only**                | Copies the text; you paste it yourself                                  | `wl-clipboard`                                                               |

<Warning>
  **Terminals usually paste with `Ctrl` `Shift` `V`,** not `Ctrl` `V`. The default mode's paste
  keystroke will not work there. Use **Type it** or **Clipboard only** when dictating into a
  terminal.
</Warning>

Your previous clipboard contents are restored after a paste by default. Turn that off in Settings
if you would rather the dictated text stay on the clipboard.

## Choosing a microphone

**Settings → Audio** lists your input devices. Sotto accepts whatever format the device provides
and converts it internally, so there is nothing to match up.

<Tip>
  A headset microphone makes a bigger difference to accuracy than a bigger model does. Whisper is
  good; it is not good enough to overcome a laptop microphone across a room.
</Tip>

## Dictating without inserting anywhere

Open the **Notepad** — from the tray menu, or `sotto --notepad` — press Record, and the formatted
text accumulates in a window with a Copy button. Useful for drafting, and for when the target
application does not accept synthesised input.

See [Notepad and CLI](/sotto/notepad-and-cli).