---
title: How do I set up Codus Whisper?
question: set up native Codus Whisper voice dictation model hotkey
area: dictation
surface: desktop
audience: beginner
keywords: [Whisper, microphone, dictation, local transcription, hotkey, model download, paste, type, sounds, history, stop and send]
related: [how-do-i-dictate-to-codus-instead-of-typing, i-dictated-but-nothing-appeared-in-the-terminal, how-do-i-set-up-superwhisper-for-dictation]
updated: 2026-09-09
visibility: public
---

# How do I set up Codus Whisper?

Open **Settings → Codus Whisper**, turn dictation **On**, download a model and choose a hotkey. Codus Whisper is built-in voice input; you do not need SuperWhisper for this workflow.

## Before you start

This guide covers the native Mac workflow. Transcription runs on your Mac after the required engine and model are available. It is not a cloud transcription subscription. Do not assume identical support on Windows; check whether your installation reports the engine ready before relying on it.

## Set it up

1. Open the **gear → See all settings → Codus Whisper**.
2. Set **Dictation** to **On**.
3. Under **Engine**, click **Install the transcription engine** if it is missing and wait for installation to complete. Under **Model**, click **Download** for your choice, wait for **downloaded**, then select it.
4. Set **Hotkey** by choosing a key or capturing a combination.
5. Under **Microphone**, allow access and choose the input device. Check **Auto-paste** for any accessibility permission needed to insert text.
6. Click **Try it** when it becomes available and speak a short test.

## Choose how it behaves

**Insert style → Paste** inserts quickly and may leave terminal paste highlighting. **Type** simulates typing and can be slower for long text. Highlighting alone is not a failure.

**Sounds** and **Sound style** control dictation cues. **Microphone buttons on Agents** can be shown or hidden independently of Chat's microphone. When shown, choose when they appear and their size.

Use normal stop to review text before sending. **Stop and send**, including the displayed shifted-hotkey action, submits it directly; use it only when you are ready to send. **History** can recover recognised text; use **Copy**, delete an individual transcript or **Clear all** when appropriate. It is stored locally and can contain sensitive text. Clearing history does not recall a message already sent.

## How do I know it worked?

**Try it** is available, the recording responds to your shortcut, and your words appear in the intended field. If not, use [dictation troubleshooting](i-dictated-but-nothing-appeared-in-the-terminal.md).