BabaVoices › Use cases › Email

Dictation for email: say the reply, paste a clean one

Most email replies are three or four sentences you already know how to say out loud. BabaVoices lets you say them, removes the ums and restarts, fixes the punctuation, and puts the finished text on your clipboard so you can paste it into Gmail, Outlook or any other mail client.

Why talking an email is faster than typing it

A typical person types around 40 words a minute and speaks at well over 100. The gap matters most for the email you answer between other things: a client asking whether Thursday still works, a colleague who needs a yes or no on a quote, a supplier chasing an invoice. You know the answer instantly, but typing it on a laptop between meetings, or with your thumbs on a phone, turns a ten-second thought into a two-minute chore. Many people put those replies off until the end of the day, and that is how a short inbox becomes a long one.

Plain dictation solves only half of this. Built-in voice typing writes down exactly what you said, including "um, so, yeah, I think, I mean, Thursday is fine," and you end up fixing it by hand. BabaVoices adds a Polish step on top of the browser's speech recognition. Polish uses Anthropic's Claude to strip filler words and false starts, join the fragments into sentences, and correct capitalization and punctuation, while keeping what you meant. The result reads like something you wrote, not a transcript.

A workflow that fits Gmail and Outlook

1. Open the email you need to answer in one tab, and BabaVoices in another (or install it to your phone's home screen).
2. Press Alt+R or tap the mic, and say the reply the way you would say it to the person across the desk. Include the greeting and the sign-off if you want them.
3. Stop. If automatic Polish is on, the cleaned version replaces the raw text; otherwise press Polish. Either way, the text is already on your clipboard.
4. Switch to Gmail or Outlook, click into the reply, and paste. Read it once, adjust a name or a date if needed, and send.

On a phone, the Share button sends the text straight into the Gmail or Outlook app, or into Messages or WhatsApp if the reply belongs there instead. For people who live in webmail on a work laptop, a BabaVoices extension for Chrome and Edge is on the way. It puts a small mic in the Gmail and Outlook compose box and inserts the polished text at the cursor, which removes the tab switch. Until it is listed in the stores, the copy-and-paste workflow above works in any browser that can dictate.

What to say for better results

Speak in whole thoughts and don't worry about stumbles; Polish handles those. Do say names and numbers clearly, because speech recognition hears "fifteen" and "fifty" as close cousins, and Polish cannot know which one you meant. If an email needs a list, say "first," "second" and "third," and the structure usually survives. For a formal message, say so in the words you use: "Dear Ms. Ortiz" will be kept, and a casual "hey" will be kept too. Polish tidies tone; it doesn't change who you are talking to.

Before and after

Here is a real-world style reply to a customer asking about a delayed order, spoken in one go while walking to a meeting.

Notice what changed and what didn't. The self-correction from Monday to Tuesday was resolved to the final answer, the repeated words and fillers are gone, and the paragraph breaks make it scannable. No new promises were added, and the apology stayed as short as it was when spoken.

Honest limitations

• You still need to read it before sending. Speech recognition can mishear names, product codes and numbers. Polish fixes grammar, not facts it never heard correctly.
• Not every browser can dictate. Firefox has no speech recognition yet. Chrome, Edge and Safari work, as do all browsers on iPhone.
• Background noise hurts accuracy. A busy café or a car with the windows down will produce more errors than a quiet room. A headset mic helps a lot.
• Long, complex emails need structure. Dictation shines for replies of one to three paragraphs. A ten-paragraph proposal is better dictated in sections, polished, and then assembled.
• Free polishes are limited. Dictation is unlimited, but the free plan includes 10 polishes a day (25 when signed in). That is enough for most people's replies on a normal day; heavy email users may want a paid plan.

Frequently asked questions