Dictate text

Recording a prompt takes one click. This page covers the control, the shortcut and the behaviour while recording.

The control

Above the prompt field you will find a microphone button labelled Dictate prompt, an animated level indicator, and a status line that starts by telling you how to begin:

Press "Dictate prompt" or use Ctrl+Shift+. to start recording.

Start recording

Select Dictate prompt, or press Ctrl+Shift+. anywhere on the page.

The first time, the browser asks for permission to use your microphone. Accept it — see Allow microphone access.

Recording then begins and the status line switches to a live counter showing elapsed time against the limit:

Recording… 0:12 / 3:00

The button stays pressed while recording, and the level indicator animates with your voice. If it does not move, the browser is not hearing you.

Stop recording

Three ways to stop:

• select the button again;
• press Ctrl+Shift+. again;
• say nothing and wait for the three-minute limit, which stops it for you.

When the limit stops the recording, the status says so explicitly:

Recording stopped at 3 minutes (limit reached).

The audio recorded up to that point is kept and transcribed. Nothing is discarded.

The keyboard shortcut

Ctrl+Shift+. toggles recording. It works from anywhere on the page, which means you can keep your hands on the keyboard while composing.

When several prompt fields exist on one screen, the shortcut targets the control that is currently recording; otherwise it picks the first available one. If you have a specific field in mind and the screen has more than one, use its button rather than the shortcut.

While it records

The button is disabled during transcription, not during recording, so you can stop whenever you like.

Speak as you would to a colleague. There is no need to dictate punctuation or to speak unnaturally slowly — see Dictation language and punctuation.

If you make a mistake mid-sentence, just correct yourself out loud and carry on. Editing the transcript afterwards is free; re-recording is not.

What happens when you stop

The status changes to:

Transcribing audio...

The recording is sent to your site and forwarded to the POM AI account service. This takes a few seconds. The button is disabled during that time so a second recording cannot start over the top of it.

On success:

Transcription added to the prompt.

The text is appended to the prompt field, and the cursor is placed at the end so you can continue typing straight away. See Review and insert dictated text.

Cost

Six credits per started minute. A twenty-second recording and a fifty-second recording both cost six credits; a two-minute-ten recording costs eighteen.

There is no estimate and no confirmation dialog — dictation transcribes as soon as you stop. The maximum exposure for one recording is eighteen credits.

The practical consequence: dictate one longer prompt rather than several short ones. Three separate ten-second recordings cost eighteen credits; the same content in one recording costs six.

Recording well

• Get close to the microphone. A laptop's built-in microphone at arm's length in a noisy office is the most common cause of poor transcripts.
• Reduce background noise. Music and conversation degrade accuracy more than accents do.
• Think first, then record. A recording of you working out what to say produces a transcript of you working out what to say.
• Use a headset if you dictate regularly. It is the single biggest quality improvement available.

If nothing happens

If the status reports that recording is unsupported, or that microphone access was blocked, start with Allow microphone access. For anything else, see Troubleshoot dictation.