If you live, work, or write in more than one language, you already know the problem: most dictation software was built for American English speakers and treats every other language as a bolt-on feature you have to pay extra for, download separately, or do without entirely.
This guide is for everyone else — the translators, the international students, the bilingual households, the expats writing emails to their parents in one language and reports for work in another. The people who pick a voice typing tool and then realise it can't actually handle their daily reality.
The short version: PeekoType uses OpenAI's Whisper model under the hood, which was trained on roughly 680,000 hours of multilingual audio. It handles 99+ languages out of the box, runs entirely offline, and costs the same whether you dictate in English, Tamil, or both in the same sentence.
What "supports 99 languages" actually means
You'll see this claim from a lot of dictation tools. Most of them mean one of two things: either they support 99 languages on paper but accuracy collapses below the top 10, or they support 99 languages but you need a live cloud connection and pay a per-minute usage fee for anything outside English.
Whisper is genuinely different. The model was trained on multilingual data simultaneously, so it doesn't have a "real" language and 98 afterthoughts. Languages with large amounts of training data — Spanish, French, German, Mandarin, Japanese, Russian, Portuguese, Italian, Korean — perform almost as well as English. Lower-resource languages still work, just with slightly lower accuracy and a tendency to need clearer pronunciation.
The languages people actually ask us about
The full list is in the app's language picker. If your language uses a non-Latin script — Chinese Hanzi, Japanese Kana/Kanji, Korean Hangul, Cyrillic, Arabic, Hebrew, Devanagari, Thai — PeekoType writes it directly in the correct script. You don't need to install a separate IME or keyboard layout; it just types the characters into whatever app you're using.
The use cases that finally got us to write this post
1. Bilingual households writing to family
You're texting your mum in Punjabi, then writing a Slack message to your colleague in English, then helping your kid with a school email in French. Switching keyboard layouts every few minutes is exhausting. With voice typing, you just switch the language in the menu — or skip the switch entirely if your sentence stays in one language at a time.
2. Translators drafting first passes
A surprising number of professional translators don't actually type their drafts — they dictate them. It's faster, less fatiguing, and forces you to read the source aloud, which helps catch awkward phrasing. Because PeekoType is 100% offline, even confidential client material stays on your machine. (If client confidentiality is your bigger concern, our guide for international teams covers data residency in more depth.)
3. International students writing essays in a second language
If you're studying abroad and writing in your second or third language, voice typing can be a huge equaliser. You think faster than you type in any language, but the gap is even wider in a non-native one. Speaking it out lets you focus on the ideas, then come back to edit the grammar — closer to how natural speech actually works.
4. Expats keeping up with both worlds
You moved abroad, your home language is now the one you use least, and you're losing fluency. Dictating a long email home once a week — at a pace where you can actually think in the language — is a small thing that adds up.
5. Language learners practising pronunciation
If Whisper transcribes what you say correctly, you can be reasonably confident a native speaker would understand you. It's not a perfect proxy for pronunciation tutoring, but it's a free feedback loop you didn't have before.
What about mixing languages mid-sentence?
This is the one place where every voice typing tool gets a little wobbly, including ours. Whisper detects the dominant language of a clip and transcribes the whole thing in that language. If you say "I'll be at the restaurante tonight", the model tries to fit "restaurante" into English or fits everything into Spanish — depending on which one it picked first.
The cleanest workaround is to dictate one language per recording chunk. Hit the hotkey, say your English sentence, release. Hit it again, say your Spanish sentence, release. It's not magic, but it's quick once it's muscle memory.
Accents, dialects, and "non-standard" speech
One of the underrated wins of training Whisper on the entire multilingual internet is that it's heard a lot of accents. We've had users report it handling:
- Glasgow, Liverpool, Newcastle, and other strong UK regional accents
- Indian English (with the rhythm and cadence intact)
- Nigerian English, South African English, Singaporean English
- Heavily accented second-language English from speakers in their 60s+
- Quebec French and Maghrebi French
- Cantonese-inflected Mandarin
It's not perfect. If you have a speech difference (stammer, dysarthria, very fast speech), expect to spend some time finding the model size and microphone setup that works best for you. But the baseline is far higher than older speech engines like the one built into Windows.
Privacy and "where does my voice actually go?"
For multilingual users this matters more than usual. A lot of cloud-based dictation services keep transcripts to train future models. If you dictate a confidential email in Mandarin to a family member who lives in mainland China, that recording is now sitting on a US server somewhere. If you're a translator working on a legal document, your client probably has a clause about where data is processed.
PeekoType runs the model locally on your Windows PC. Your audio never leaves your machine. There's no telemetry on what you said or in what language. For anyone who needs to take data residency seriously — translators, lawyers, healthcare workers, journalists, NGO staff — that's the actual selling point.
How to actually get started (the 90-second version)
- Install PeekoType on your Windows PC.
- Open settings, pick your default language (you can change it any time).
- Choose the model size: small is faster and works well for the major languages, medium or large is better for low-resource languages and heavy accents.
- Press the hotkey (F9 by default), speak, release.
- That's it.
If you've never used voice typing before and want a more detailed walk-through, see our 5-minute beginner setup guide.
The cost question
This is where multilingual users get hit the hardest by traditional dictation software. Dragon Professional doesn't support most languages at all on the consumer version. Otter.ai's free tier limits non-English use sharply. Google Docs Voice Typing is cloud-only. Every cloud-based service that does handle multiple languages charges per minute, per month, or per language pack.
PeekoType is £19.99 once, for every language Whisper supports, for as long as your computer keeps running. That's the entire pricing model.