Daily use

Dictate with your voice · speak instead of typing

There is a single microphone, anchored to the field you are in, and it disappears when the focus leaves. Click it or press Ctrl+Shift+D, speak, and the text lands at the cursor. Once the transcription arrives, "Clean up with AI" shows up as a second, separate action that only replaces the dictated passage. The audio is deleted as soon as it becomes text.

Writing up a visit, filling in a ticket note or answering a long email costs far more time typed than spoken. Voice dictation exists for that part of the day: you speak, and the text lands in the field exactly as you said it.

One microphone, on the field you are in

There is no microphone icon on every field. There is one button, anchored to the field that has FOCUS, and it disappears when the focus leaves. People who never dictate never see it; people who do have the microphone exactly where they are writing. If the page scrolls and the field leaves the window, the button hides with it instead of floating loose.

How to dictate

Click the field you want to write in. The microphone appears at its top right corner.
Click the microphone, or press Ctrl+Shift+D without taking your hands off the keyboard (on a Mac, ⌘+Shift+D works too).
The first time, the browser asks for permission to use the microphone. Allow it.
Speak. While it records, the button turns red and the label changes to "Stop and transcribe".
Click the microphone again, or press Esc, to end the recording.
Wait for "Transcribing…". When it finishes, the text lands in the field.

💡 The text lands AT THE CURSOR, not at the end of the field. Dictate in the middle of a paragraph you already wrote and the passage goes where you left the cursor. A space is added only when there is text before it that does not already end in a space, so dictating twice in a row neither glues the sentences together nor opens a gap at the start of the field.

Where dictation works

Text inputs and text areas across the system: a record's form, a comment, a ticket note, the body of an email in the inbox.
It also covers fields that do not exist yet: the microphone follows the focus, so no screen has to know dictation exists in order to offer it.
Leaving the field does not cancel a recording in progress, so you can click the button itself without losing what you are saying.

Where it deliberately does not appear

Password, number and date fields: dictating there produces junk, so the microphone is not offered.
Read-only or disabled fields.
Rich text editors, such as the one in the Documents module and the marketing flow email editor. They are not ordinary text fields and do not accept text written from the outside.
The terminal, which has a button of its own, described below.

"Clean up with AI" is a second action

What lands in the field is always what you SAID, raw. Once the transcription arrives, a "Clean up with AI" button appears next to the microphone: it adds punctuation, capitalization and paragraph breaks and fixes obvious speech-to-text slips, without adding, removing or reinterpreting content, and without answering the text. It only runs when you click it.

The cleanup replaces ONLY the dictated passage. Whatever you had typed in the same field before stays exactly as it was. If you would rather keep the spoken wording, just dismiss the notice with the X.

In the terminal

The terminal has a microphone of its own, at the bottom right of the window, because there is no ordinary field there: the terminal draws its own prompt. The button behaves the same way, with one difference that is a safety decision: dictation is NOT sent on its own. The transcribed sentence lands on the command line and waits for you to read it and press Enter. In the terminal a sentence becomes an executed action, and sending it by itself would make the machine act on a misheard word.

What happens to the audio

The recorded audio goes straight to the system's storage, under your company's own space, is read once to be transcribed and is then DELETED. That includes when the transcription fails: failing is no reason for the file to stay. What remains is the text in the field, plus an audit trail entry saying a dictation happened and how many seconds it lasted.

Before you start: transcription has to be switched on

Dictation uses the SAME transcription engine as call and meeting recordings. If transcription is not configured and enabled for your company, the microphone simply does not appear: offering a button that will not work is worse than not offering it.

Ask an administrator to open Settings → Integrations and find the "Call transcription (speech-to-text)" card.
Pick the provider and enter the account key.
Save, then click "Enable".
Reload the screen: the microphone starts showing up on text fields.

Dictating consumes transcription, it is not a free extra. If your company brought its own speech-to-text account, those seconds are billed by that provider. When transcription runs on the system key, the seconds are metered and land on your company's Sellio bill, exactly like the ones from call and meeting recordings. "Clean up with AI" is an AI call and lands in AI metering, which is why it only runs on a click.

Limits and notices

There is an 8 MB audio ceiling per dictation. Beyond that it is not dictation, it is a recording: for meetings, use meeting transcription.
Microphone permission belongs to the browser. If you deny it, the product says so once ("Microphone access was denied. Allow it in your browser to dictate.") and does not insist; re-allow it in the site permissions to dictate again.
If the recording carries no speech at all, the message is "Nothing was heard." and nothing is written to the field.
Opening and closing without saying anything costs no transcription: audio that is too short is discarded before it is uploaded.
The recording state is also announced to screen readers, so people using one know when the microphone is open.

Example: A field technician at Delta Parts finishes a visit and has to log what she did. She opens the ticket, clicks the note field, presses Ctrl+Shift+D and talks for a minute about what she found, what she replaced and what is still pending. She presses Esc, waits for "Transcribing…", and the write-up appears in the field. Because she wants the paragraph to read well for the customer, she clicks "Clean up with AI": punctuation arrives, the text gains paragraphs, and nothing she said changes meaning. What she had already typed at the top of the note is untouched.

💡 Which language is recognized depends on the transcription provider configured for your company. If you work in more than one language, it is worth testing a short dictation in each before you rely on it day to day.

Open this article inside the system →