Auto captions: generate subtitles from a video's own audio
How DijiFlow Dictate turns a video's soundtrack into timed subtitle lines on your computer, and which of its four out...
You have %itemCount% in your cart.Total being %total%
Dictate in 29+ languages with DijiFlow, transcribed 100% on your device. No cloud processing, no account. About 98% accuracy, and no internet needed to transcribe after one download.
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 its speech models on your own machine, so you can dictate in 29+ languages without your audio leaving your device.
That one decision — keep the model local — is what makes multilingual dictation actually practical. Your audio is not uploaded, 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.
It's worth being precise about the numbers, because they mean two different things. DijiFlow Dictate recognises 90+ languages with auto-detect, and that is the widest range it can turn into text. Within that range, 29 languages have full support: you can select them by name, and on Pro and the trial they get a layer of vocabulary tuning, so domain terms, names, and the phrasing you actually use come through cleaner instead of being guessed at. Outside the 29, languages are not tuned, and the text may come out in English.
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.
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 and no upload: your voice and the text it becomes stay on the machine in front of you.
On-device transcription also strips out the friction that quietly drags multilingual work down:
Key takeaway
Because every language runs on-device, transcription keeps working without a connection once its model is downloaded — mid-flight, off-grid, or anywhere the Wi-Fi gives out.
The trade-offs line up cleanly once you put them side by side:
| Capability | DijiFlow (on-device) | Cloud dictation |
|---|---|---|
| Audio stays on your machine | ✓ | ✗ |
| Transcribes without internet | ✓ | ✗ |
| Switch languages mid-sentence | ✓ | ✗ |
| No account or per-minute billing | ✓ | ✗ |
The Mac app itself is small, around 34 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, transcription needs no internet. On Pro and the trial, Large is the most accurate, Express is close to it and faster, and Fast is the quickest, in 16 languages including English; the free version uses Standard for English and Enhanced for other languages. You choose where to sit on that line.
Day to day, DijiFlow Dictate lives in your menu bar on Mac (the system tray on Windows) 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.
On-device multilingual dictation pays off most for people who move between languages as a normal part of the workday:
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 stays on your machine is what you've been missing, take a look at how DijiFlow Dictate fits the way you work.