The floating panel
The floating panel is the small panel that floats above your work while you dictate. It shows a waveform that moves with your actual voice, the transcription forming in real time underneath it, and the pause, stop, and cancel buttons.
It comes in three layouts: expanded, compact, and mini. You can drag it anywhere, and when you stop talking the wave winds into a spinning ring while the app works on your words.
Expanded, compact, or mini
The panel comes in three layouts, and you pick the one you want in Settings. Expanded shows the live transcript. Compact is one row. Mini is just the waveform. A button on the panel steps through the three sizes. Its position and colors are yours to set as well.
Drag the panel anywhere; it remembers its place.
With VoiceOver, every state of the panel is announced, and its controls are exposed as accessibility actions.
The buttons and the keys
The circular buttons are the same everywhere in the app: pause and resume, stop and insert, and cancel. Stop ends the dictation and types your words at the cursor. Cancel throws the dictation away; nothing is typed or saved.
You do not have to click them. While you are dictating, Space pauses and resumes, Esc cancels, and 1 through 9 switch the output mode for that dictation. The order matches the Modes list, so 1 is always the first mode in your list. Space and the number keys are only captured while you are actually dictating, so they never interfere with normal typing.
Right Command starts and stops a dictation from any app. It works two ways, and you pick in Settings. In toggle mode, one tap starts and the next tap stops. In hold mode it becomes a walkie-talkie: recording runs while the key is down and stops the moment you let go. Any key can be a trigger, not just modifiers.
The waveform and the ring
The waveform is real. It moves with your actual voice and rises the instant you start talking. The text underneath it is the transcription forming in real time.
When you stop, the wave winds into a spinning ring while the app thinks. That is the app working on your words: the speech model turns your voice into text, then the cleanup model polishes it if your mode uses one, and then the panel closes. A long recording does not leave you on the ring. It streams out as you go, cut at natural pauses.
The first dictation after launch loads the model, which takes a moment, and the panel tells you while it happens. The speech model warms up at startup and the cleanup model loads on its first use. Every dictation after that is quick.
If the panel says Didn't catch that, the recognizer heard silence. Get closer to the microphone, check the input source in Settings, or raise the input boost if you speak quietly.
Colors, and turning it off
The pop-up and the waveform have their own colors, separate from the app's accent color, and there is a full RGB mode.
If you would rather have no panel at all, turn it off in Settings. Dictation still works exactly the same. The capybara in the menu bar still tells you what is happening: its speech bubble turns red and breathes while recording, becomes a spinner while transcribing, and flashes green when your text was inserted.
Questions
Can I turn off the YapToText floating panel?
Yes. Turn it off in Settings. Dictation still works exactly the same, and the capybara in the menu bar still shows a red speech bubble while recording and a spinner while transcribing.
Why did the panel turn into a spinning ring?
The ring means the app is working on your words: transcribing them, then cleaning them up if your mode uses the cleanup model. A long recording streams out as you go instead of leaving you on the ring.
How do I move the floating panel?
Drag it anywhere; it remembers its place. Its position, size, and colors are all yours to set.
How do I switch between the expanded, compact, and mini panel layouts?
Pick the layout you want in Settings: expanded shows the live transcript, compact is one row, and mini is just the waveform. A button on the panel also steps through the three sizes.
The panel said Didn't catch that. What happened?
The recognizer heard silence. Get closer to the microphone, check the input source in Settings, or raise the input boost if you speak quietly.