Google voice typing alternative: dictate anywhere

What Google voice typing does well, and where it stops

Credit first. Google voice typing is free with a Google account, needs zero installation, and works in the latest versions of Chrome, Edge, and Safari. Inside a Docs document, or speaker notes and captions in Slides, the recognition is decent and the language list is long. If you draft everything in Docs, it earns its keep, and our Google Docs voice typing walkthrough covers the setup in detail.

The stops are just as concrete. Punctuation is manual: you say "period," "comma," "question mark," "new paragraph." And punctuation is not available in every language. The editing and formatting commands work only in English, and only when both your account language and the document language are English. (Slides speaker notes get voice typing but no voice commands at all. I do not know why either.)

And the boundary that started this article: the feature exists inside two Google surfaces and nowhere else. Gmail, Slack, your CRM, your code editor, a web form — silence. None of this is a bug. It is a free feature doing what its own help page says it does, no more. The question is whether your writing fits inside that box. And when it stops listening mid-sentence, that is usually the microphone, not you — our Google Docs voice typing fixes cover the usual causes.

A voice typing alternative that works in every app

Whisper takes the opposite bet: dictation belongs to the operating system, not to one website. You press a hotkey: Ctrl+Space on Windows, or Command+Option held down on a Mac as push-to-talk. You speak. You release. The text lands wherever your cursor is: an email, a chat window, a spreadsheet cell, a commit message. There is no copy step, because there is no "source document" to copy from.

The speed argument is simple arithmetic. Dictation runs at about 145 words per minute against around 40 for typing — call it a 3.6x difference. On an M1 MacBook Air with the small English model, the gap between releasing the key and seeing text is 1.4 seconds.

Under the hood you pick one of three paths, and the app does not pick for you. Cloud mode connects to OpenAI with your own API key and uses gpt-4o-mini-transcribe or gpt-4o-transcribe. Parakeet, NVIDIA's local engine, runs 5-10x faster than Whisper models on CPU and covers English plus 24 European languages. And the local open-source Whisper models range from a 140 MB English model to the 3 GB Large v3. The app ships for Windows and for Apple Silicon Macs.

Google voice typing vs Whisper, side by side

The criteria, picked before the table: where the tool types, what it runs in, punctuation, voice commands, language coverage, offline behavior, and price shape. Those are the seven things that decided it for me when I still used the Docs microphone. The Google column comes from Google's own help page; the Whisper column comes from the app's source code, which I can see from here.

The verdict matches the first paragraph: Whisper is the pick the moment your writing leaves Google Docs. Google voice typing remains the right answer for people whose writing never does. That case gets its own section below, and I mean it.

Voice typing online vs. voice typing on your desktop

"Voice typing online" most often means a website you open in a tab: a free page with a microphone button that transcribes into a text box. These tools sit on the browser's speech recognition engine, and MDN's Web Speech API documentation is blunt about how that works: "Your audio is sent to a web service for recognition processing, so it won't work offline." The same page names the two costs: privacy, because your speech leaves the machine, and reliability, because no connection means no transcription.

For a one-off note, an online voice typing page is fine. The tax shows up with repetition: every dictation ends with select, copy, switch, paste. I once dictated a grocery list into the wrong tab and pasted it into a budgeting spreadsheet. For about ten seconds, Q3 projections included "yogurt the small ones, not the big ones." A desktop tool removes that whole class of accident because the text starts where it ends — at your cursor. And when the browser microphone misbehaves, the fix lives in browser settings, which is its own afternoon. We wrote up the Chrome dictation fixes after enough people asked.

Punctuation and cleanup: the editing tax nobody counts

The boring truth is that transcription was never the slow part. The cleanup is. With Google voice typing you speak your punctuation ("period," "new paragraph"), which works, but turns you into a person who narrates commas. I tested the commands for this article, and my test document ended up containing the word "comma" eleven times. The habit follows you home, too: I have since ended a spoken sentence, in a real conversation, with "period."

Whisper routes the raw transcript through an AI cleanup pass instead: grammar fixed, filler words removed, punctuation placed where it belongs. In local mode that runs through Ollama; in cloud mode it uses OpenAI models, with gpt-5-mini as the default enhancer. Start a sentence with "Hey whisper" and the AI treats what follows as an instruction instead of text to type: "Hey whisper, make this reply polite but firm" is a real workflow, not a demo. The difference in feel is hard to overstate. One tool asks you to perform the formatting. The other lets you talk like a person and hands back something you would send as-is.

Privacy: where your voice goes

Here is the opinion I will defend: dictation that ships your voice to a server you do not control is a privacy problem, and for some work it is a dealbreaker. Browser-based voice typing works by sending audio out for processing. That is fine for a grocery list. It is a different conversation for a draft contract, a patient note, or the email about your kid's school situation.

I watched the cost side of this lesson up close. A team I worked with had a contractor build an internal AI dictation prototype that called a cloud API for every utterance. End of quarter, the manager opened the cost dashboard: a five-figure bill, most of it one team transcribing standup recordings four times because the retry logic was too eager. The contractor suggested optimizing the prompt. The CFO suggested not paying a server to listen to meetings that already had notes. Both were right about something.

Whisper's local mode closes the question instead of arguing it: the model runs on your machine, nothing leaves the device, and no internet is needed at all. Same hotkey, same flow, zero audio in transit.

When to skip Whisper and stay with Google

If every word you dictate lands in a Google Doc, stay with Google voice typing. It costs nothing, it is already in the Tools menu, and inside Docs it does the job — that is a hard combination to argue with, and I will not pretend otherwise. The same goes if you dictate twice a month: installing a desktop app to save two copy-pastes is over-engineering, and I say that professionally. Whisper earns its place when your writing regularly leaves the document (email, chat, forms, notes), or when punctuation-by-AI, offline use, or keeping audio on your machine matters. Until then, the free built-in tool is the right tool.

That teacher email from the top of the page got sent, eventually — dictated into Gmail itself, no relay through a Doc, punctuation included without me saying a word of it. The tab with Google Docs stayed closed. That is the whole comparison in one moment: the free tool asks your writing to come to it, and the alternative goes where the writing already is. Pick whichever direction matches your day.