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%
The story and product principles behind DijiFlow Dictate: private voice-to-text, with your voice and text kept on your computer.
Open almost any dictation app and the same thing happens out of sight: you speak, and your voice leaves the room. It travels to a server, gets transcribed by a model you cannot see, and comes back as text a moment later. The convenience is genuine. But so is the trade you just made: your most unguarded form of input, recorded and processed on hardware you do not own.
DijiFlow Dictate started from a stubborn question: why should that be the default? It is voice-to-text with 100% on-device transcription, built for people who write, code or take notes and would rather their voice stayed with them. This page is less about us and more about what we are trying to prove.
Good software should earn your trust by design, not ask you to take it on faith. If your voice never leaves your device, privacy stops being a promise and becomes a fact.
The decision at the heart of DijiFlow Dictate is easy to say and hard to engineer: transcription runs locally. Speech recognition happens on your own machine, with speech models that run on the device itself. Your audio and text are not uploaded. No account and no cloud processing.
That is not a feature we bolted on for a marketing line. It is the whole point. When the audio never leaves your computer, there is no server log to leak, no breach that can expose your transcripts, and no quiet policy change that starts sending your words somewhere new. The architecture does the work that privacy policies usually have to. We would rather build the guarantee than write it.
It also means transcription keeps working when the network does not. You can pull the plug on your connection mid-sentence and dictation carries on exactly the same. That is the bar we hold ourselves to: if a tool needs the cloud to do its main job, it is not truly yours.
A handful of principles shape nearly every decision we make about DijiFlow Dictate.
The safest data is the data that was never collected. We design so the sensitive thing, your voice, simply never travels anywhere it could be lost.
Dictation should feel like an extension of typing, not a place you visit. The app lives in the menu bar on Mac and in the system tray on Windows; you press a hotkey, speak, and the text lands where your cursor already is.
We tell you plainly what each speech model costs in size and what it buys in accuracy, and we keep a genuinely useful Free tier instead of crippling it to force an upgrade.
We stand on open foundations rather than reinventing them. DijiFlow Dictate is built on EasyDictate by Charles Lukowski, used under its MIT license, and we are glad to say so.
We did not invent speech recognition, and we are not going to pretend otherwise. DijiFlow Dictate is made by PITIR TECH and grew out of open-source work, including Whisper, WhisperKit, and EasyDictate by Charles Lukowski, whose MIT-licensed code is part of what we built on. Crediting that is not a footnote we tolerate; it is part of how we think the field should work. The interesting problem was never the model alone. It was making something private, fast, and pleasant enough that you reach for it without thinking, on a Mac today and on Windows too.
The DijiFlow team is small and focused. We are not building a platform that monetizes your data, and we are not interested in collecting more than the product needs to function, and your voice and your text are never part of that. We do not want your email to start dictating, an ad profile built around you, or your recordings to train the next model. None of that is on the table, and your recordings never reach us.
The measure of success we actually care about is quiet. It is DijiFlow Dictate becoming part of how you work without demanding your attention, so you never have to wonder where your voice went. The answer is always the same, and it is the entire reason we built this. It stayed with you.
If that way of building resonates, the best way to understand it is to see how people put it to work. Take a look at the ways DijiFlow Dictate fits into real work.