DiktaSupport
Dikta is an offline dictation app. Hold the hotkey, speak, release, and the text is inserted wherever your cursor is. Everything is transcribed on your own machine. If something is not working, the answers below cover the common cases.
Contact
Email [email protected]. Include your operating system version, the dictation language you were using, and the quality tier if you know it. Replies usually come within two working days.
System requirements
- Windows 10 version 2004 or later. Windows 11 recommended.
- macOS on Apple silicon.
- A microphone.
- Around 4 GB of free memory for the highest quality tier. The faster tiers run comfortably on much less.
- An internet connection once, to download the speech model. After that Dikta works with no network at all.
Getting started
On first run Dikta asks for your interface language, your dictation language, and a hotkey, then downloads the speech model for the language you picked. The download is between 0.5 and 1.5 GB depending on the quality tier, and only happens once per language.
After that: put your cursor in any text field, hold the hotkey, speak, and release. The text appears where the cursor is. You never need to open the Dikta window to dictate.
Troubleshooting
Nothing happens when I press the hotkey
Check that Dikta is running — it lives in the system tray on Windows and the menu bar on macOS, not the taskbar. If the chosen hotkey is already claimed by another application, pick a different one in Settings.
The microphone is not picking anything up
On Windows, open Settings → Privacy & security → Microphone and confirm microphone access is allowed. On macOS, System Settings → Privacy & Security → Microphone. If you have more than one input device, select the right one in Dikta's settings — device numbering can change between sessions when devices are plugged in or removed.
The model download failed or stalled
Models are downloaded from Hugging Face. If the download was interrupted, restart Dikta and it resumes. A corporate proxy or firewall that blocks huggingface.co will prevent the first download; every later use is offline.
Transcription is slow
Speed is governed by the quality tier. Try a faster tier in Settings — on a typical laptop processor the fastest tier handles a short sentence in around a second, the default tier in around three. Dikta measures your machine during setup and suggests a tier, but you can override it at any time.
The app is slow to start, or fails to load a model
The highest quality tier needs roughly 4 GB of free memory. On a machine that is short on memory it can fail to load. Dikta refuses to promote itself to that tier automatically under 4 GB free, but if you selected it manually, drop down a tier.
Some words come out wrong every time
Add them to the custom dictionary in Settings — names, colleagues, company terms and jargon. Dikta biases the model towards those spellings and corrects near-misses.
Log files
If you report a problem, the log helps. Settings has an Open log folder button. The files are at:
- Windows:
%LOCALAPPDATA%\dikta\logs\dikta.log - macOS:
~/Library/Caches/Dikta/logs/dikta.log
Logs record what the app did, not what you dictated.
Your data
Audio and transcripts never leave your computer. Settings, dictionary and history are stored locally:
- Windows:
%APPDATA%\dikta - macOS:
~/Library/Application Support/Dikta
History can be exported or cleared from within the app. Full detail on the privacy page.
Refunds and licence
Dikta is a one-time purchase, not a subscription. Purchases, refunds and licences are handled by the store you bought from — the Microsoft Store on Windows, the App Store on macOS — under that store's own refund policy. Dikta has no account system and stores no payment details.