Dictation software FAQ

Dictation software captures audio from a microphone, converts speech into text with a speech-to-text model, then places that text where you are typing. WhisperJot is a desktop dictation app: press a hotkey, speak, and the words are typed into the focused app. Transcription runs on your device by default with the local engines. The opt-in Jot Cloud engine is available if you would rather not run a local model.

See also: Browser demo, Use cases

Most people speak faster than they type, so dictation can cut the time spent on first drafts, email, and notes. Speed still depends on how you edit: you may fix names, punctuation, or structure after the text lands. WhisperJot is built for that loop — press a hotkey, speak, and keep working in the app that already has focus rather than copying from a separate transcription window.

See also: Use cases, Dictation guides

Dictation is speaking in real time so words appear in the document or field you are using. Transcription converts recorded audio into text after the fact, often as a file you review later. WhisperJot is a dictation app that types into the focused desktop app. It also has a meetings mode that transcribes what your microphone hears into a timestamped transcript stored locally — it does not join calls or write summaries.

See also: Meeting notes, Use cases

Modern dictation uses neural speech-to-text models rather than simple command lists. WhisperJot runs those models in two ways. Jot Local and Jot Local Pro transcribe on your computer by default. Jot Cloud is an opt-in engine that sends dictation audio to our servers only while you have it selected, and that audio is never stored there. Optional text cleanup can also run locally on 127.0.0.1.

See also: System requirements, Compare

Accuracy varies with microphone quality, background noise, accent, and the speech model. Clear English in a quiet room is the usual strong case; overlapping speakers and heavy noise are harder. WhisperJot offers Jot Local for fast, English-focused dictation and Jot Local Pro for higher on-device accuracy on longer speech. Review names and numbers before you send important text. We do not publish a single accuracy score.

See also: System requirements, Compare

Jot Local and Jot Local Pro are the default on-device engines. Jot Local is fast and English-focused for short utterances; Jot Local Pro is the higher-accuracy local engine for longer speech. Both run on your computer and work offline after the model download. Jot Cloud is the opt-in cloud engine: it skips the local model footprint, runs only while selected, and never stores audio on our servers. You can switch engines anytime.

See also: System requirements, Pricing

For the local engines, yes — on first dictation WhisperJot downloads a speech model to your device. Jot Local is about 500 MB on a Mac (around 600 MB on disk on Windows and Linux); Jot Local Pro offers several models from about 80 MB up to 1.5 GB depending on the one you pick. After that, local dictation runs offline. The opt-in Jot Cloud engine downloads no local speech model and transcribes only while you have it selected. The app itself is a separate download from the Download page.

See also: System requirements, Download

Local engines need enough RAM and disk for a speech model — about 8 GB of memory and roughly 3 GB of free disk at minimum, more if you enable local LLM cleanup. If your machine is older or low on memory, use the opt-in Jot Cloud engine: it has no local model download or RAM floor, needs an internet connection, runs only while selected, and never stores audio on our servers. Same $99/year plan.

See also: System requirements, Pricing

English is fully supported. The language picker also offers auto-detect plus Spanish, French, German, Italian, Portuguese, Dutch, Russian, Polish, Chinese, Japanese, Korean, Hindi, Arabic and Turkish. For the best results outside English choose the Jot Local Pro engine or the opt-in Jot Cloud engine; the fast Jot Local engine is English-focused, and the automatic text-cleanup rules are written for English, so turn cleanup down when dictating in other languages.

See also: System requirements, Use cases

Press the global hotkey, speak, then release it or press again to stop — hold for push-to-talk, or tap to toggle. The words are typed into whichever desktop app currently has focus. You can change the hotkey in Settings. Grant microphone access and, on a Mac, Accessibility (plus Input Monitoring if you use a keyboard shortcut) so the app can hear you and insert text. A 30-second browser demo shows the idea before you install.

See also: Browser demo, Download

WhisperJot types into the focused desktop app, so it works in editors, browsers, email, chat, terminals, and notes apps — no plugin required. It is not a phone keyboard and does not inject into other computers. Per-app settings can change how text is inserted. If an app does not accept pasted or injected text, focus a normal text field and try again.

See also: Dictation guides, Use cases

Yes. Custom vocabulary lets you teach WhisperJot names, product terms, and jargon so they are spelled the way you expect. Add entries in the app; they apply on that computer. Shortcuts can also expand a spoken phrase into longer text. This is spelling correction after recognition, not a promise about domain accuracy — still review drafts that matter before you send them.

See also: Use cases, Dictation guides

Yes. WhisperJot applies automatic punctuation and capitalization so spoken sentences land as readable text, and it strips filler words. You can also say voice commands such as period, comma, question mark, new paragraph, and all caps to shape the draft while you talk. Automatic cleanup rules are written for English, so turn cleanup down when dictating in other languages.

See also: Dictation guides, Use cases

WhisperJot has a meetings mode that transcribes only what your microphone can hear, on-device, with no bot joining the call and no summaries. It is meant for in-person meetings, interviews, lectures, and your side of a call — not for capturing the other participants' system audio. Export a timestamped transcript as Markdown or plain text. Meeting notes are included in the $99/year plan.

See also: Meeting notes, Use cases