Help Centre
Questions,
answered.
Everything about DijiFlow Dictate — setup, privacy, features, pricing, and quick fixes.
Getting started
Download the app from the download page, open the file, and drag DijiFlow into your Applications folder — the usual Mac install. The first time you launch it, the app lives in your menu bar (not the Dock) and walks you through granting the two permissions it needs and downloading your first speech model. The whole thing takes a couple of minutes, and there's no account to create.
By default you dictate with Right Option + Space. Hold it to talk (push-to-talk) and release to insert your text, or tap it to toggle recording on and off — whichever feels natural. The hotkey works in any app, anywhere on your computer. You can remap it to a different combination in settings, and Pro lets you set up to three separate slots for different languages or shortcuts.
DijiFlow needs two macOS permissions. Microphone lets it record while you dictate — and the mic is only ever live while you are actively recording. Accessibility is used for a single purpose: detecting your global dictation hotkey (Right Option + Space) so push-to-talk and toggle work from any app, and inserting the finished text where your cursor is. It does not read, monitor, or log what you type elsewhere, and nothing you type is recorded or sent anywhere.
The speech model is the part that actually turns your voice into text, and it runs entirely on your computer. To keep the app tiny (~5 MB) we download the model the first time you use it rather than bundling every option. This one-time download needs an internet connection; after that, dictation and file transcription work completely offline. Pro can fetch the larger, most accurate models, including Large-v3.
Privacy & security
Nowhere. Your audio and the resulting text stay on your computer and are never uploaded. The microphone is active only while you are recording, and DijiFlow then inserts the text straight into whatever app you are typing in. There is no telemetry, no analytics, and no crash reporting inside the app, and no account is required to use it.
Yes. Because everything runs on-device, the privacy guarantee is verifiable rather than a promise: once a speech model is downloaded, dictation and file transcription need no network access at all. You can confirm this yourself — turn off Wi-Fi (or your whole network connection) and watch transcription keep working. Nothing leaves the device, so there is simply no server-side record of your usage to keep.
No. The app contains no telemetry, no analytics, and no crash reporting — it doesn't phone home. You don't create an account and you don't sign in. (Separately, the DijiFlow website is a normal Shopify store and does use cookies and analytics to process orders and measure site traffic; that's entirely independent of the app, which uses neither.)
Very little, and all of it local. Activation stores only a hash of your licence key and a hash of a hardware ID, kept on your computer — there's no server-side profile and no record of what you transcribe. The hardware ID is an opaque identifier, not your name or email, and it's only used to confirm the licence is being used on a reasonable number of devices.
It's a strong fit. Because audio and text never leave your computer, there's no third party processing your recordings and no cloud copy to secure — which suits sensitive work such as legal, medical, and other confidential dictation where data residency and GDPR-style obligations matter. We can't make a formal certification claim on your behalf, but the on-device, no-account, no-telemetry design removes the usual cloud-transcription privacy risks. For organisation-wide deployment, see our company licensing options.
Features
Two ways to dictate, your choice. Push-to-talk: hold the hotkey while you speak and release when you're done — great for short bursts and quick replies. Toggle: tap once to start, tap again to stop — better for longer passages where you don't want to hold a key. Both insert the finished text wherever your cursor is, in any app.
Yes. DijiFlow dictates into whatever app has your cursor — email, chat, documents, code editors, browser fields, notes, anywhere you can type. It isn't tied to one program. Because it lives in the menu bar and listens for a global hotkey, you can dictate without switching windows or opening the app first.
Yes. As well as live dictation, DijiFlow transcribes existing audio and video files entirely on your computer — no upload. The Free tier handles files up to 3 minutes; Pro removes that limit for unlimited file and video transcription, so you can run long recordings, interviews, and meetings locally.
Yes, on Pro. Pro adds speaker labels (who said what) and timestamps to transcripts, which makes interviews, meetings, and multi-person recordings far easier to read and reference. Both run on-device like everything else. These features aren't included on the Free tier.
Yes, on Pro. Add your own custom terms and use vocabulary packs so names, brands, product names, and industry jargon transcribe correctly instead of being guessed phonetically. This makes a big difference for medical, legal, and technical work where specialised vocabulary is common.
29 languages, with automatic multilingual model selection so you don't have to pick one manually — DijiFlow chooses the right model for what it hears. All 29 languages are available on the Free tier. Pro adds the largest, most accurate models (including Large-v3) for the best results on accents, technical terms, and longer recordings.
Accuracy & models
DijiFlow runs OpenAI's Whisper speech-recognition models through WhisperKit and Apple's CoreML, so transcription happens on your Mac's Metal GPU and Neural Engine rather than a remote server. The result is fast, accurate, fully on-device transcription with no cloud round-trip.
On clean speech, accuracy is around 98%. Real-world results depend on your microphone, background noise, accent, and the model you choose — a larger model is more accurate for difficult audio. On Pro you can add custom vocabulary so names and technical terms come out right, which lifts accuracy further for specialised work.
Many times faster than real time. With Pro's GPU and Neural Engine acceleration, DijiFlow processes audio several times quicker than it would take to play it back, so live dictation feels instant and long file transcriptions finish quickly. The Free tier uses CPU processing, which is slower but still fully functional.
For everyday dictation, a smaller model is fast and accurate enough, and it's the default. If you need the highest accuracy — for strong accents, technical or medical terms, or long recordings — choose a larger model. Pro unlocks the largest models, including Large-v3, alongside GPU and Neural Engine acceleration that keeps even the big models responsive.
Platforms
Windows is coming soon. DijiFlow Dictate is available on Mac today, and the Windows release — supporting AMD, Intel, and NVIDIA GPUs — is next. Leave your email on the download page and we'll notify you the moment it's ready. Pro licences are intended to cover Windows when it launches, so a purchase today isn't Mac-only forever.
DijiFlow runs on macOS 14 (Sonoma) or later on Apple Silicon (M1 and newer). It relies on the Metal GPU and Neural Engine for on-device transcription, so Apple Silicon is required — Intel Macs aren't supported. The app itself is a small ~5 MB download. Speech models are separate and download once on first use: the base model is about 300 MB, and the largest (Large-v3) is around 6 GB. DijiFlow lives in the menu bar, ready whenever you press your hotkey.
Pricing & licensing
Free (free forever) gives you core push-to-talk and toggle dictation, all 29 languages, one language/hotkey slot, file transcription up to 3 minutes, and CPU processing. Pro adds three language/hotkey slots, unlimited file & video transcription, speaker labels, timestamps, vocabulary packs and custom terms, GPU/Neural Engine acceleration, and the largest, most accurate models (including Large-v3). See the full breakdown on the features page.
The trial gives you 30 days of full Pro features with no credit card required — every Pro capability, including the largest models, GPU/Neural Engine acceleration, unlimited file and video transcription, speaker labels, and timestamps. When the 30 days end, the app simply reverts to the Free tier (free forever), so you keep core dictation, all 29 languages, and short file transcription. Just download DijiFlow to begin — the trial starts on first launch.
There are four Pro plans: Monthly £4.99, 6-Month £20.94, 12-Month £29.94 (50% off, our best value), and Lifetime £39 one-time. The app is identical across every paid plan — they differ only in how long the licence stays valid. Larger organisations can also get custom Enterprise pricing. See full details on the pricing page.
It depends on how long you plan to use DijiFlow. Monthly (£4.99) is the lowest commitment and easy to cancel — good for trying Pro in real work. The 12-Month plan (£29.94, 50% off, our best value) suits anyone settling in for the year. Lifetime (£39 one-time) pays for itself in well under a year and never renews. Roughly: if you expect to use it past about eight months, Lifetime is the better deal. A 6-month option (£20.94) sits in between.
Yes. The Lifetime plan is a single £39 payment with no renewals — your Pro licence stays active for good. It's the best long-term value if you use DijiFlow regularly, and because Pro licences are intended to cover the upcoming Windows build too, a Lifetime purchase isn't tied to just one platform forever.
Yes. Updates are included for as long as your licence is valid — on Lifetime that means forever, with no extra charge. You don't pay separately for new features or improvements; an active Pro licence keeps you on the latest version.
Yes — there's a 30-day refund. If DijiFlow isn't right for you within 30 days of purchase, email info@pitirtech.com and we'll sort it out. We reply within one business day, Monday to Friday, 9:00–18:00 UK. Full details are on our refund policy page.
Yes. If you're buying for a team or an organisation, we offer bulk and Enterprise licensing with custom pricing — ideal for privacy-conscious teams that need on-device transcription at scale. Visit our company licensing page or email info@pitirtech.com and we'll put a quote together.
Troubleshooting
A few things usually help:
- On Pro, enable GPU/Neural Engine acceleration — it's many times faster than CPU.
- Use a smaller model for everyday dictation; reserve the largest models for difficult or critical audio.
- Close other heavy apps so the GPU and Neural Engine are free.
The Free tier runs on CPU, which is slower by design — upgrading to Pro is the biggest single speed boost.
Try these in order:
- Pick a larger, more accurate model for difficult or noisy audio.
- Add your own terms via custom vocabulary and vocabulary packs (Pro) so names, brands, and jargon transcribe correctly.
- Reduce background noise and speak at a steady, natural pace.
- Check your microphone is selected correctly in macOS sound settings.
Still stuck? Email info@pitirtech.com and we'll help — usually within one business day.
Open DijiFlow from the menu bar, choose the licence or activation option, and paste in the key from your order confirmation email (keys look like PTIR-XXXX-XXXX-XXXX-XXXX). Activation stores only a hash of the key and a hash of a hardware ID locally — no account sign-in is needed. If a key won't activate, double-check it matches your order, then contact info@pitirtech.com and we'll get you running.
No problem. Your key is in your original order confirmation email, so check that first (including spam). If you can't find it, email info@pitirtech.com from the address you ordered with and we'll re-send it. We reply within one business day, Monday to Friday, 9:00–18:00 UK.
We're happy to help.
We reply personally, within one business day (Mon–Fri 9:00–18:00 UK).
Free forever · No account · 100% on-device