Speech to text for ESL learners

Why writing English by voice helps when it's not your first language

The hard part of writing in a second language is rarely the ideas. It's the friction between the idea and the page. You know what you want to say. You can say it out loud. But typing it means fighting spelling, word order, and a keyboard layout while the sentence you had in mind quietly evaporates. Speaking skips most of that fight.

Dictation throughput sits around 145 words a minute, against roughly 40 for typing. For a native speaker that's a nice speedup. For someone composing in their second or third language, it's bigger than that, because typing in a non-native language is slower and more error-prone to begin with. You spend the saved effort on the part that matters — saying it right — instead of on mechanics.

And there's a quieter benefit. When you speak a sentence and see it written back, you find out fast whether it actually makes sense. A clumsy phrase looks clumsy on the screen in a way it doesn't always sound in your head. That feedback loop is useful for a learner, the same way it helps a student turning a lecture into notes. It won't correct your grammar by itself — that's the AI pass, two sections down — but seeing your own words on the page is its own small lesson.

Press a hotkey, speak English, the text lands at your cursor

The mechanic is simple, which is the whole appeal. You press a hotkey, you speak, you release, and the transcript pastes at your cursor in whatever text field has focus. Whisper holds a short tail after you let go of the key, so your last word doesn't get clipped — handy when you're still finding the end of an English sentence. Because it pastes at the OS cursor, it works the same in your email, a Google Doc, a chat box, or a homework assignment.

There's nothing to wire up per app. No browser extension, no plugin, no token to paste. Your cursor is in the box, you talk, the words appear. A small capsule shows up while you speak so you know it's listening:

The hotkey is worth setting once and forgetting. On Windows it's Ctrl+Space; on Mac it's Command+Option, a modifier-only push-to-talk you hold while speaking. Both are changeable in Settings if they clash with something you already use. If you've ever set up voice typing in Google Docs, this is the same idea, except it isn't trapped inside one app — the same key fills every box on your screen.

Set it up in two minutes (Windows or Mac)

You need a Mac on Apple Silicon or a Windows 10-or-newer PC, a working microphone, and any app you want to write into. The whole local pipeline is free for any signed-in account, with no payment method asked for at sign-up. Here's the sequence.

The slow part is the model download, not the setup. Everything else is the four steps above. Once it's running, writing an English email stops being a typing task and starts being a talking task — which, when English is the part you're still practising, is the part you want to keep.

voice to text on Windows · on Mac

How well it handles accents, and the 99-language trick

Let me be straight about accents, because this is where the honest answer matters. Whisper's models were trained on a wide range of real-world speech, including a lot of non-native English. In practice that means a strong but non-native accent is usually transcribed accurately. Usually. Heavier accents, fast speech, background noise, or a name and a technical term in the same sentence will still trip it up sometimes. It's good. It is not magic, and anyone telling you it gets every accent perfectly is selling you a demo, not a Tuesday.

Two practical levers move the needle more than the model choice does. First, the microphone — a $20 USB mic does more for accuracy than any model upgrade, full stop. Speaking clearly and a touch slower helps too, which is no hardship when you're practising English anyway. Second, the model family. The multilingual Whisper builds cover 99 languages and tend to handle accented English better than the English-only builds, because they've heard far more of the world. Local Parakeet covers English plus 24 other European languages — 25 in total — and is the fastest local option, but it can't translate. The English-only .en builds are exactly that: English only, no translation.

That last point opens a genuinely useful trick for learners. The multilingual Whisper models can translate to English as they transcribe. So when an English sentence won't come — the word is on the tip of your tongue in your first language but gone in English — you can say it in your native language and get an English draft back. It's not a polished translation, and I wouldn't ship it untouched, but as a way to draft in your strongest language and then refine in English, it removes a real wall. Say it in the language you think in; fix it in the language you're learning.

Local or cloud: which mode fits a language learner

For most ESL writing — emails, homework, messages, a first draft of an essay — local mode is plenty, and it's free and offline. Cloud earns its place when you want top accuracy on a hard recording or you need to look something up mid-sentence. Here's how the three paths differ, because the app makes you pick and I'd rather you pick well.

The split comes down to speed, language coverage, and where your voice goes.

• Local Parakeet — NVIDIA's TDT engine, around 600 MB, and the fastest local option — 5 to 10 times faster than Whisper on CPU. Covers English plus 24 other European languages, 25 in total. No translate-to-English. If your first language is European and you're writing in English, this is the quick, fully offline pick.
• Local Whisper — slower than Parakeet on the same machine, but the multilingual builds cover 99 languages, tend to handle accents better, and can translate to English. Pick this for Chinese, Japanese, Korean, Arabic, or any language Parakeet can't do, and for the draft-in-your-language trick. Default English model is around 480 MB.
• Cloud (OpenAI, BYOK) — best accuracy and web access, using your own OpenAI key billed straight by OpenAI. Transcription runs on gpt-4o-mini-transcribe by default. Needs internet, so it's the one path that leaves your machine. The Cloud surface is part of Whisper Pro.

The boring truth is that for everyday English writing, the local multilingual Whisper model covers most learners well: 99 languages, decent accent tolerance, translate-to-English when you need it, and nothing sent to a server. Both local engines run fully on your machine, which matters if you're dictating anything you'd rather keep private — a personal essay, a job application, a message you're nervous about getting right. Start local. Reach for cloud only when local leaves you wanting more accuracy.

Turning a spoken draft into clean written English

Raw dictation comes out as a run-on, in any language. You say a sentence the way you'd say it out loud — with a filler word, a self-correction, a comma you didn't voice — and that's the unpunctuated wall any speech engine hands back. For a learner this is where speech to text earns its keep, because the cleanup pass does the part that's hardest in a second language.

Whisper can run an AI pass over the raw text before it lands. Say the activation phrase "Hey whisper" and the spoken draft gets enhanced — filler words stripped, punctuation added, the run-on split into sentences, obvious slips smoothed out. On a local model that runs through Ollama; in cloud mode it's gpt-5-mini by default. It tidies grammar and structure rather than rewriting your meaning, so the result still sounds like you, just cleaner.

One honest limit. The AI pass fixes grammar and shape; it does not teach you why it changed something. If your goal is to learn the rule, read the before-and-after side by side — the diff is the lesson. If your goal is just to get a clean message out the door before a deadline, let it clean and move on. Both are fine uses; they're different goals, and only you know which one you're after today.

That same speak-then-clean flow works for everything you write, not just one app — you can type faster with your voice across email, docs, and chat, so a long paragraph becomes a few spoken sentences instead of a wall you type out one careful word at a time.

When speech to text is the wrong tool for an ESL learner

I'd rather lose your click than waste your time, so here's where a dictation tool is the wrong answer. If your goal is to improve your pronunciation, this is not it. Speech to text turns your speech into text; it does not score your accent, correct how you say a word, or tell you that "thirty" came out as "dirty." For that you want a language app built for pronunciation, or a tutor, or a conversation partner. A transcription tool is a writing aid, not a speaking coach, and pretending otherwise would be dishonest.

A few more honest off-ramps. If you only need to dictate a short message, the free tools already on your machine cover it — on Windows, press Windows key + H for the built-in Voice Typing bar; on Mac, turn on Dictation in System Settings under Keyboard, and on Apple Silicon general text can be processed on-device. The Windows one needs an internet connection and routes through Microsoft's servers, so it isn't offline. And if you want a tool to actually teach you grammar rules with explanations and exercises, that's a grammar checker or a learning app — the AI cleanup here fixes the text, it doesn't run a lesson.

Reach for a dedicated, system-wide dictation tool when the writing itself is the bottleneck: long emails, essays, applications, anything where you can say it faster than you can type it in English, and you want one hotkey that behaves the same in every app on Windows and Mac. Below that bar, use what's free, or use the right tool for the job. The right call sometimes points away from us, and I'll always say so.

If you're choosing where to dictate, the platform guides cover the setup in detail — voice to text on Windows walks through the same flow step by step on a PC.

English is my third language, and I wrote most of this guide by speaking it into a text box, then letting the cleanup pass fix the seams I'd never catch by ear. That's the honest pitch: it won't make your English perfect, and it won't teach you the rules, but it will get the sentence out of your head and onto the page far faster than your fingers can. The fixing is still yours. The fast part is the help.