Your first dictation in 60 seconds

Onboarding, step by step

On first launch Dimmy runs a short wizard: welcome, pick a model, set your shortcut, try it. Here is what each step asks for.

‹›

Pick a model

Choose where transcription runs. Both platforms now show two cards: Local and Cloud (Groq).

• Local: private, offline, runs on your machine. Pick a model from the dropdown (Parakeet, around 466 MB, or a Whisper size), then download it. You see a live progress bar with a percent while it downloads. If it fails you get a Download failed message with the reason and a Try again button. The Parakeet bundle resumes a partial download instead of starting over on both platforms; Whisper sizes also resume on Windows, but on macOS a Whisper retry re-downloads from scratch.
• Cloud (Groq): fastest and most accurate. Paste a Groq key in the gsk_... field. Dimmy validates it on the spot: Validating…, then Ready to use, or the error if the key is rejected. No key yet? Use the Get API key link, which opens console.groq.com/keys.

Set your shortcut

Pick the keys you dictate with. The step starts from the shortcut Dimmy is already using, so if you re-run setup it shows your real combo, not a default.

• Windows: a free capture box. Click it, then press and hold the combo you want. Dimmy records exactly those keys, Windows key included. Choose Push-to-Talk or Toggle below. The combo you set is applied to the live hotkey before the Try it step, so the keys you just chose are the keys that fire.
• Mac: the default is the fn key. The step offers presets fn, Control+Option, and Control+Shift, plus a Custom recorder. You can change it later in Settings under Shortcut.

Try it

Dimmy shows the pill and arms your shortcut so you can test recording right inside the wizard before you finish.

The 3 steps

1. 1Click into any text field. A note, a browser tab, a chat box, a code editor.
2. 2Press your hotkey. On Mac that's the fn key by default, on Windows and Linux it's Win + Alt. Speak naturally.
3. 3Release the hotkey. The pill goes through transcribe → enhance → done, and the text appears where your cursor was.

Watch the pill

The pill in the corner of your screen tells you what is happening. Idle when waiting, rainbow border with a live waveform while recording, dots while transcribing, a coloured border during AI enhancement, a green check when text was pasted.

If nothing pastes

• Make sure a text field was focused before you pressed the hotkey.
• On macOS, grant accessibility permission (System Settings → Privacy → Accessibility → Dimmy).
• If the cursor moved away while you were speaking, the text still lands in the clipboard. Cmd/Ctrl + V to paste it manually.

Redo the setup

Want to run the wizard again? Open Settings, go to the About page, find Run setup again and click Open setup. The wizard reopens and is idempotent: a model you already downloaded and a key you already saved are detected and kept, so nothing is re-downloaded and you are not asked for a key twice. Same on Windows and Mac.