Dictate in any app · AI chat
ChatGPT has a microphone, but voice mode is a conversation — not text you can edit before sending. Typally types what you say straight into ChatGPT — hold a key, talk, let go — so you can dictate anywhere ChatGPT lets you type.
This is a vendor guide from the makers of Typally. Anything we say about ChatGPT is worth verifying — third-party features change without notice.A dictation button and a full voice mode. Voice mode is a spoken conversation, which is a different activity from composing a careful prompt: you cannot see the text, reread it, or fix the middle of it before it is sent.
A good prompt is a written artifact. You want to say it, then look at it, then adjust it. Voice conversation skips the looking-at-it step, and the dictation button lives inside one website rather than everywhere you work.
In practice people use it for: Long prompts with real context, pasting in a problem and describing it properly, follow-up questions that build on an answer, and drafting the same prompt for several different tools.
Nothing changes inside ChatGPT itself. Typally sits in the tray and types into the focused window:
One install, one sign-in. Nothing else to configure.
The cursor position is the whole instruction — ChatGPT routes the keystrokes from there.
Hold, talk, release. Long passages are fine.
Enter sends the message in ChatGPT. If your dictation ends with a return, the prompt goes before you have read it.
Typally types into the composer so you can read the prompt before sending it, which voice mode does not let you do. It records for as long as you hold the key with no cut-off, so a full paragraph of context is no harder than a sentence. The cleanup pass removes the ums, so the model reads a clean prompt rather than a transcript.
A marketer wants feedback on a positioning idea. Typed, the prompt is two sentences and gets a generic answer. Spoken, it becomes a paragraph covering the product, the audience, what has already been tried and what she is worried about — then she reads it back, fixes one word, and sends. Same effort, far better answer.
Dictate the prompt into the box, read it once, then send. That extra second of reading catches the transcription slip that would otherwise send the model in the wrong direction.
Enter sends the message in ChatGPT. If your dictation ends with a return, the prompt goes before you have read it.
Voice mode is a spoken conversation with no editable text. Dictation puts words in the input box so you can read and change them before sending.
A system-level tool types into whichever window has focus, so the same hotkey works across every AI chat, plus your editor and inbox.
More relevant context generally helps. Speaking removes the typing cost that pushes people toward short prompts.
Free plan: 100 minutes of dictation a month, no card. Pro is $7.99/month if you want more.
Also dictate in: Cursor · VS Code · Gmail · Microsoft Copilot · NotebookLM · see all apps