Get started
Your first transcript
Run guided setup, download a model, and turn your first recording into text.
On this page
Start with a short recording using a local speech model. Once the basics work, you can add AI cleanup, walkthroughs, meetings, or a remote engine.
1. Install and open OpenWhisper#
Follow Installation for your platform. On macOS, approve the first launch and allow Microphone access.
2. Run guided setup#
A brand-new install offers guided setup a moment after the main window appears. It takes three short steps, and every choice saves as you make it:
- Record: your microphone, with Test to check that it hears you (nothing is recorded); your recording shortcut, and whether you press it to start and stop or hold it to talk; and the speech engine. If its model isn't on the computer yet, Download fetches the model and the runtime it needs right there, with a progress bar and a retry if it fails; a model bundled with an all-in-one installer shows Included and ready. With an English-only engine, Other languages switches to a multilingual one.
- Clean up: optional AI cleanup, where the text would be sent, and whether a key is needed. Setup never turns cleanup on by itself.
- Output: auto-paste, the clipboard, and the live transcript.
Click the step tabs to move in any order; Enter continues. Setup ends on a Ready page with a box where you can try your shortcut straight away; it tells you whether the dictation worked, or why not. If something is still missing, such as the speech model or an API key for cleanup, the page is titled Almost ready and lists it with a button to fix it. Not now dismisses setup for good; File → Guided setup… runs it again whenever you like. Existing installs aren't offered it, but can open it from the same menu.
3. Get to know Settings#
Settings opens on Home: where your words go, the everyday controls (microphone, shortcut, voice model, cleanup, auto-paste, theme, updates), and what's downloaded. Every card links to its full page. The sidebar groups the rest into six sections, and the search box in the header finds any setting by name.
| Section | Pages |
|---|---|
| Dictation | Recording, Voice model, AI cleanup, Remote engine, Walkthroughs |
| Personalize | Dictionary, Snippets, Styles, Learned rules, Profiles, Commands |
| Automate | Actions, Hooks, API & MCP |
| Meeting Mode | Voice & speakers, Intelligence, Fast judgments, After the meeting, Dashboard |
| Models & storage | Downloads, Runtime, Backup & restore |
| App | General, Appearance, Hotkeys, API keys, Developer options |
These docs write paths as Settings → Section → Page, for example Settings → Dictation → Voice model.
4. Choose and download a speech engine#
Open Settings → Dictation → Voice model. New installs default to Parakeet Unified (English) on Windows x64 and Linux x86_64, Parakeet MLX on Apple Silicon with macOS 14 or newer, and Local Whisper elsewhere. Existing installs keep their previous choice.
Keep the default for this first recording, or compare alternatives in Choosing a model. A local engine needs no API key.
Then open Settings → Models & storage → Downloads and install the selected model and its runtime. Wait for both to finish before recording. The Windows and Linux packages already include Parakeet Unified's runtime, and an all-in-one install already has a model marked Included.
5. Select your microphone#
Open Settings → Dictation → Recording and choose the microphone to use. You can rank several; OpenWhisper falls back to the next one if your first choice disappears.
6. Record a few sentences#
In Quick Record, press the record button, say a sentence or two, and stop. OpenWhisper transcribes when you stop; supported engines also show a live preview while you speak. New installs using Parakeet Unified or Parakeet MLX start with live preview on.
For example, say: “This is my first OpenWhisper recording. I am checking that my microphone works.”
7. Use the finished text#
Copy the result from Quick Record. Dictation follows your clipboard and auto-paste settings in Settings → App → General; on macOS, auto-paste also needs Accessibility permission. Before relying on auto-paste, try a recording in a text editor and leave it focused while the recording finishes.
Make it your own#
- Set a recording shortcut in Settings → App → Hotkeys, and choose Toggle or Push and hold.
- Turn on AI cleanup for punctuation, spelling rules, or output profiles.
- Explain something on screen with a walkthrough.
- Use Upload File for recordings you already have; see Dictation & audio files.
- Pick the overlay's look and the theme in Settings → App → Appearance. New installs start with Bars. Live preview layout chooses whether the look shrinks into a slim header above live words or stays full size below them.
- Going offline? Download everything first; see Privacy & offline.
If you get silence, a model error, or no pasted text, Troubleshooting walks through each step.
Get help from the community, or tell us what went wrong.