Voice and dictation on the iPhone
Tap the mic, talk as long as you like, and pick when the words appear.
On the iPhone you can talk wherever you can type: the chat bar with the operator, the New task row, a title field, the body of a task or a note. One recorder sits behind all of them, a take can be any length, and nothing is ever cut short. Settings → AI and Dictation → Dictation decides whether the words appear as you speak (Live) or land in one piece once you stop (After recording, the default, and the more accurate of the two).
Where the mic is
The chat bar. On the Operator tab the resting bar reads "Ask your operator" with a mic at its end. Tap the mic and recording starts in the same instant, with no keyboard. The bar shows a capsule: an X on the left, the waveform, and an up arrow on the right. The up arrow ends the take and sends it to your operator in one move. The X drops the take and clears the bar. Typing during a take cancels the take and keeps your draft. A take over a draft adds the words after it.
A field with a keyboard. The New task row and a sheet's field show a small bar above the keyboard with a dismiss circle and a mic. Tap the mic and it grows into the same capsule. The up arrow stops the take and the words land in the field where the caret was. The X takes them back out, and a take that heard nothing leaves the field as it was.
The body of a task or a note. The mic sits at the end of the body's formatting bar. The words land at the caret, and you can keep editing while it records.
From outside the app. The Action Button and the widget's mic open Roma recording into the chat bar.
Talk to Roma with your voice describes what the operator does with a spoken message on every client.
Live, or after recording
Settings → AI and Dictation → Dictation has two choices. After recording, the default: nothing appears while you talk, the capsule's bars dim and sweep while the words land, and then the whole take arrives at once. Live: the words appear as you say them, in the chat bar or the field, and are tidied as the take goes on. The page says it plainly: live is less accurate than after recording, because after recording hears the whole take at once. The choice belongs to this phone, and every phone starts on After recording.
Long takes, calls and headphones
A short take, under about thirty seconds, is transcribed the moment you stop. A longer one is cut into pieces and transcribed as you go, so stopping is quick even after twenty minutes of talking. The take keeps recording when you leave the app: the Dynamic Island shows Roma's icon with the clock, and a press and hold on it shows the X and Send.
A call, Siri or an alarm that takes the microphone pauses the take, and it carries on when the microphone comes back. Headphones connected or removed mid-take are picked up without a break. Everything heard so far is kept whatever happens, and a take the service could not transcribe live is transcribed from the recording instead.
What happens to the words
In the chat bar, the words go to your operator as said, and the operator does with them what it does with anything you say. In a task or a note, the dictated text is shaped by Settings → AI and Dictation's Level and Voice, the way AI editing describes, and the New task row's words become the task's title and due date, the way Create a task describes.
Questions people ask
Does dictation work without a connection?
No. Transcription runs on Roma's service, so a take needs the network. Typing works without one, and what you type syncs when you are back.
Why do the words appear only after I stop talking?
Because Dictation is set to After recording, the default. It is the more accurate choice. If you would rather watch the words arrive, choose Live under Settings → AI and Dictation.
Is there a limit on how long I can talk?
No. A take can be a one-line task or half an hour of thinking out loud, and nothing is truncated: a long take is transcribed in pieces as you speak.
Can I look at the words before they are sent?
In the chat bar the up arrow sends the take in one move, so the words never wait in the draft. In a field they land in the field for you to edit. And with Send after silence on the Action Button, an auto-stopped take waits three seconds for you to call it off. Settings on the iPhone lists the switches.