120+ languages
English, Arabic, Hindi, Japanese, Spanish, Persian, Russian, Turkish — every language the recognition service supports, searchable in Settings and switchable in a click.
Version 5.2.2 · 7-day free trial · macOS 14+
Speak any language. Your Mac types it.
Hold a key, say what you mean, let go — the words appear wherever your cursor already is. 120+ languages, including every one macOS dictation leaves out.
Apple silicon · 2.9 MB · $9/month after the trial
Persian سرور رو روی پورت سه هزار دیپلوی کن
server رو روی port 3000 deploy کن
🌐
fn
Done — typed at the cursor
Listening …
Features
Small app, careful details
Everything here exists because dictating in a second language kept breaking in the same handful of ways.
Your shortcut, your way
Hold a key while you speak, or press once to start and again to stop. Bind fn, a function key, or any combination you like.
Four ways to transcribe
Start with no account and no key at all. Run Whisper entirely on your Mac, or bring an OpenAI or Gemini key. Keys live in your keychain.
Borrowed words stay readable
Non-Latin scripts transliterate English technical terms. AnyDict puts them back — say “database” in Persian, Arabic or Russian and that is what gets typed.
Numbers become digits
Spoken numbers are written as digits in every language that has a table, including compound forms. Latin, Persian or Arabic numerals, your choice.
Stays on your Mac
Pick an on-device model and the audio never leaves your Mac at all. On the cloud providers it is sent only while transcribing, and the recording is deleted the moment it is done. Recent transcripts stay local and clear whenever you like. No account, no telemetry.
Rewriting, in practice
Recognition writes everything in the language's own script. AnyDict puts borrowed technical vocabulary back into Latin letters and converts spoken numbers — per language, and only where it should.
Persian سرور رو روی پورت سه هزار دیپلوی کن server رو روی port 3000 deploy کن
Arabic الكود فيه باغ وعشرون خطأ الcode فيه bug و20 خطأ
Russian деплой на сервер, двадцать пять минут deploy на server, 25 минут
Providers
Pick where your voice goes
Nowhere at all, or the service you already pay for. The default needs no account; a local model needs no network.
Google Speech
no keyThe default. No account and no key of your own. Shared quota of roughly 50 requests a day, best on clips under 15 seconds.
OpenAI
your keyWhisper and the GPT-4o transcription models. Strong on long, accented and mixed-language speech. Billed per minute.
Google Gemini
your keyTranscribes audio natively and is generous at low volume. Particularly good on non-English speech. Billed per token.
On your Mac
offlineWhisper running locally on the Neural Engine. Download a model once and the audio never leaves the machine. No key, no quota, and it works on a plane.
Keys are stored in your login keychain, never in plain text, and are sent only to the provider you chose.
How it works
Three seconds to learn
There is no interface to master. The whole app is one key.
- 01
Pick your language
Any of 120+, from Settings. The vocabulary and number handling adapt to it.
- 02
Hold your key and speak
The menu bar icon fills in while it listens. Release when you are done talking.
- 03
Your words appear
About a second later the text is typed at the cursor, wherever that happens to be.
Languages
120+ languages and dialects
Twenty flavours of Spanish, fifteen of Arabic, and everything in between. Pick one in Settings and the whole app adapts to it.
🇬🇧 English 🇪🇸 Spanish 🇸🇦 Arabic 🇮🇳 Hindi 🇮🇷 Persian 🇫🇷 French 🇩🇪 German 🇵🇹 Portuguese 🇷🇺 Russian 🇯🇵 Japanese 🇰🇷 Korean 🇨🇳 Chinese 🇹🇷 Turkish 🇮🇹 Italian 🇳🇱 Dutch 🇵🇱 Polish 🇻🇳 Vietnamese 🇹🇭 Thai
🇮🇩 Indonesian 🇸🇪 Swedish 🇬🇷 Greek 🇮🇱 Hebrew 🇧🇩 Bengali 🇱🇰 Tamil 🇵🇰 Urdu 🇰🇪 Swahili 🇺🇦 Ukrainian 🇨🇿 Czech 🇷🇴 Romanian 🇲🇾 Malay 🇵🇭 Filipino 🇳🇵 Nepali 🇩🇰 Danish 🇫🇮 Finnish 🇳🇴 Norwegian 🇭🇺 Hungarian
🇧🇬 Bulgarian 🇭🇷 Croatian 🇷🇸 Serbian 🇸🇰 Slovak 🇭🇰 Cantonese 🇿🇦 Afrikaans 🇪🇹 Amharic 🇦🇲 Armenian 🇦🇿 Azerbaijani 🇬🇪 Georgian 🇰🇭 Khmer 🇱🇦 Lao 🇱🇰 Sinhala 🇮🇳 Telugu 🇮🇩 Javanese 🇿🇦 Zulu 🇮🇸 Icelandic 🇱🇹 Lithuanian
+ 70 more, searchable in Settings
What it is not
On the default provider, AnyDict uses the same speech endpoint Chromium uses for the Web Speech API. That means no key and no account of your own, but it is not an official product and comes with real limits worth knowing before you rely on it.
- Roughly 50 requests a day on the default provider. The key is shared, so the practical ceiling moves around.
- Short utterances work best there — up to about 10–15 seconds. Recording stops at 30.
- An OpenAI or Gemini key lifts both limits , at their usual per-minute or per-token pricing.
- Not notarized yet. On first launch, right-click the app and choose Open.
Dictate in your language
7-day free trial, then $9/month. 2.9 MB, and it stays out of your way.
Download AnyDict 5.2.2 Contact
Get in touch
Bug reports, a language that needs its own term rules, or just to say it worked — all welcome.