Dictate your message instead of typing it

The thing people often search for is voice typing on any website. This is narrower. Dictation in Copilotly types into the Copilotly chat box, and from there you can send the message, or copy the answer where you want it. It does not type into the fields of other websites.

To work on text that is already in another website's field, use the selection toolbar described on rewrite anywhere: select the text and ask for a change, including with a question typed in the command menu.

Dictation relies on the browser's own speech recognition (the Web Speech API), not on a recogniser we host. Copilotly shows the microphone button only if your browser provides it, so nobody sees a button that cannot work. In practice that means it is available in Chromium-based browsers such as Chrome and Edge, and absent where the browser does not offer the API, as in Firefox and some Safari setups.

The browser decides how audio is processed. In some browsers speech recognition is done by a cloud service run by the browser vendor, so check your browser's privacy documentation if you dictate anything sensitive. Copilotly receives the text of your message, not the audio.

It listens in your browser's language by default (the language set in your browser). If your browser is set to English and you speak Spanish, recognition quality will suffer.

• Say punctuation clearly or add it afterwards; recognisers differ in how much they insert on their own.
• Dictate the question, not the polished answer. Then use chat to tidy it, or ask for a tone and length.
• Use a decent microphone in a quiet place, and allow microphone access when asked. If you denied it, the chat shows Allow microphone access for this site to dictate; change that in the site settings of your browser.
• Dictate the first draft, then fix it. Raw speech has false starts. Ask the chat to clean it up.

Two different things share the word. Dictation turns your speech into text in the chat box. My Voice is a writing-style profile learned from samples you paste, used when Copilotly drafts replies in Gmail, LinkedIn, X, WhatsApp, Teams and Craft. They are covered separately on memory and personalization.

My Voice has a daily limit: the free plan uses your voice on 5 drafts a day, and Pro uses it on every draft. We do not list a Copilotly daily limit on dictation itself; the message you send is a normal chat message. Premium-model replies count against your allowance as usual, and your own API key is never metered.

Speaking is faster than typing for most people, and the difference shows up on the jobs where you would otherwise stop to think about wording.

• Long questions with background. Describing a situation out loud, including the awkward details, usually gives the model more to work with than a terse typed line.
• Brain dumps. Say everything you know about a topic in one go, then ask the chat to organise it into headings, a plan or an email. The email writer turns a rough set of facts into a draft the same way.
• Accessibility and comfort. If typing is slow or uncomfortable, dictation lets you use chat with far less strain.
• Away from a keyboard. A tablet or a laptop on your knees is easier to speak to than to type on.

It is less useful for exact strings: code, email addresses, URLs, passwords or product codes. Recognisers get those wrong. Type them.

A workable routine: dictate a rough version, read it once for mishearings (names and technical words are the usual casualties), send it, and ask for the answer in the form you want. If you then want to turn the answer into text in another site, copy it, or open the selection toolbar in the field and tidy it there.

If your browser does not offer speech recognition, your operating system almost certainly has its own dictation, and the Copilotly chat box accepts what it types just like anything else.

The first time you dictate, the browser asks whether the site may use your microphone. Allow it, or the button will not record.

If you clicked Block by mistake

The chat shows Allow microphone access for this site to dictate. Open your browser's site settings for the page you are on, set Microphone to Allow and try again. Chrome's own help page on using your microphone in Chrome shows where to find it.

If the wrong microphone is used

The browser chooses the default input device of your operating system. Change the default in your system sound settings, or in the browser's microphone setting, and dictate again.

Language and accents

Recognition uses your browser's language, as described in the MDN documentation for the lang property. Dictating a language your browser is not set to gives poor results. Change the browser language first.

Because Copilotly dictation fills the chat box only, people who want voice typing in Gmail, Google Docs or a form need a different tool.

Your operating system

Windows, macOS, iOS and Android all include system dictation that types into any focused text field. It works in every site and every browser, and it is usually the most reliable option for general voice typing.

The browser or the web app's own tool

Some web apps, such as Google Docs, have voice typing built in. Check the app's menus first.

Dictate in Copilotly, then move the text

The practical hybrid is to dictate a rough message into the chat, ask for a clean version in the tone you want, and copy the result into the other site. Tools like the email writer and the tone changer do the same cleanup without an account. For editing text already in a field, use rewrite anywhere.

Dictation is built on the browser's speech recognition, and what happens to the audio is the browser's decision.

Where the audio goes

In some browsers the audio is sent to a cloud service run by the browser's vendor for transcription. In others it is processed on the device. The MDN overview of the Web Speech API explains that the browser, not the website, supplies the recogniser, and that availability varies.

What Copilotly receives

Copilotly receives the text that ends up in the message box when you send it. It does not receive the audio. If you want to review the text before it leaves your browser, dictation gives you that chance, since the words appear in the box first.

What to do with sensitive material

If the content is confidential, such as client details or health information, either type it or use an Incognito chat, and check your browser's privacy documentation for how speech is processed. A good rule is to dictate the shape of the question and leave out names and numbers.

A few habits turn mediocre recognition into messages you are happy to send.

Before you speak

Open the sidebar with Alt+C, check the microphone indicator in your browser and think of the first sentence. Starting with the question helps both you and the model.

While you speak

Speak at an even pace, a few words at a time, and pause between thoughts. Say numbers and dates clearly. If you say a name the recogniser does not know, spell it afterwards in the box.

After you stop

Read the box once. Fix names, numbers, product terms and anything that changed meaning. Then send. If the text is messy, send it anyway and add: "Clean up the wording of my message above and then answer it."

Use the answer where you need it

Copy the reply into your document or email. If the reply is for another person, ask for a tone and length in the same message, or adjust it later with rewrite anywhere. For a deeper question that needs sources, say it out loud and choose deep research or web search before sending.