# Voice in Redcast

> Dictation into any field, cloud modes, spoken questions and voice edits.

Redcast has separate voice workflows. Choose the one matching what you want to happen when you finish speaking.

| Workflow | Result | Setup |
| --- | --- | --- |
| [On-device dictation](/docs/voice/dictation) | Your words typed at the cursor | Download a local recognition model |
| [Cloud dictation and modes](/docs/voice/cloud) | Transcribed, polished or translated text | Approve an audio-capable connection |
| [Ask by Voice](/docs/voice/ask) | An answer about the window in front | AI, Screen Awareness and an approved voice model |
| [Edit Selection by Voice](/docs/voice/ask#edit-selected-text) | A rewrite of selected text | Select text and speak an editing instruction |
| [Dictation History](/docs/voice/history) | Find, copy, paste or retry previous dictations | Choose a retention period |

## Permissions

Recording needs **Microphone** access. Typing into other apps and reading selections need **Accessibility**. A window screenshot requires a separate **Screen Recording** grant. Voice features are off until you enable them.

## Choose a shortcut deliberately

Dictation supports toggle and hold-to-talk bindings. Assign separate shortcuts to dictation and Ask by Voice so you know whether your words will be typed or treated as a question. Check Settings → Shortcuts for conflicts with macOS and other Redcast commands.

## Where audio goes

On-device dictation transcribes locally. Cloud dictation and voice questions send the recording only to the connection you approve. A changed or removed connection does not silently redirect audio to another provider.

Successful dictation audio is discarded. Failed dictation recordings can be kept temporarily for retry; Ask by Voice and voice edits never keep their audio. Transcripts and answers have separate [history settings](/docs/voice/history).