Type in 90+ Languages, Fully Offline: Multilingual Dictation Without the Cloud

Switch languages mid-sentence and most dictation tools stumble. They send every word to a remote server, lock you into one language per session, and go quiet the moment your connection does. DijiFlow Dictate works the other way around: it runs OpenAI's Whisper speech models on your own machine through WhisperKit and CoreML, so you can dictate in 90+ languages without a single byte of audio leaving your device.

That one decision — keep the model local — is what makes multilingual dictation actually practical. Nothing uploads, nothing waits on the network, and there are no cloud minutes to meter. You press a hotkey, speak, and the words land at your cursor in whatever app you happen to be in.

• 90+
  languages transcribed on-device
• 29
  with dedicated vocabulary tuning
• 100%
  offline after the model downloads

90+ languages transcribed, 29 tuned for your words

It's worth being precise about the numbers, because they mean two different things. The Whisper large-v3 model transcribes 90+ languages — that's the full range DijiFlow Dictate can turn into text. Within that range, 29 languages get an extra layer of vocabulary tuning, so domain terms, names, and the phrasing you actually use come through cleaner instead of being guessed at.

So the breadth is wide and the depth is real. Dictate a paragraph in Spanish, drop in an English technical term, finish a note in German — all from the same setup, no mode switch, no menu. On clear speech, accuracy lands around ~98%, and dictation runs several times faster than typing. That gap widens fast when the alternative is hunting across a keyboard layout you don't use every day.

Why offline is the point, not a footnote

Working across languages usually means working with material you'd rather not hand to a third party: client correspondence, legal text, medical notes, research interviews, internal documents. The moment that audio goes to a cloud service, it becomes someone else's question to answer. On-device transcription removes the question entirely. No account, no upload, no telemetry — your voice and the text it becomes never leave the machine in front of you.

Offline also strips out the friction that quietly drags multilingual work down:

• No round-trip latency. Transcription runs locally, so nothing waits on a distant server to answer.
• No connection required. On a plane, in a basement office, on hotel Wi-Fi that drops every few minutes — dictation just keeps working.
• No per-minute meter. Speak as much as you want, in any language, without watching a usage counter.

On-device multilingual vs. the cloud

The trade-offs line up cleanly once you put them side by side:

What it costs in space, what it gives back in speed

The app itself is small — around 12 MB. The speech models are separate and download once, ranging from roughly 300 MB to 6 GB depending on which one you pick. After that first download, everything runs fully offline. Larger models handle accents, mixed vocabulary, and less common languages more confidently; smaller ones stay lighter on resources. You choose where to sit on that line.

Day to day, DijiFlow Dictate lives in your menu bar and stays out of the way. There's no window to manage and no document to copy out of. Email clients, code editors, chat apps, word processors — if you can type into it, you can dictate into it.

Who gets the most out of it

On-device multilingual dictation pays off most for people who move between languages as a normal part of the workday:

• Translators and interpreters drafting and reviewing across language pairs.
• Researchers and journalists transcribing interviews and notes without exposing the source material.
• Bilingual professionals writing to colleagues and clients in more than one language every day.
• Anyone privacy-minded who would simply rather their speech not touch a remote server.

Try it in the languages you actually use

DijiFlow Dictate has a Free tier that stays free, a 30-day Trial with the complete feature set, and a Pro option you can buy monthly, for 6 or 12 months, or once for a lifetime license. The honest test is to install it, start the Trial, and dictate in your real languages — that's the only way to feel how on-device transcription fits your workflow.

If multilingual dictation that never leaves your machine is what you've been missing, take a look at how DijiFlow Dictate fits the way you work.