Ask and edit by voice

Ask by Voice combines a spoken question with context from the window in front. Enable AI and Screen Awareness, then enable Ask by Voice and approve the connection that will hear audio. It has its own shortcut, separate from Dictation.

Ask about a windowCopy Anchor Link

1. Put the relevant app in front.
2. Press the Ask by Voice shortcut and speak your question.
3. Finish recording. Redcast sends the question and captured context to the approved model.
4. Read the streamed answer in the floating result panel.

Try “Explain the error in this window” or “Summarize the selected paragraph.” If screenshots are enabled and permitted, the model can receive the window image as well as readable text.

Some requests ask for text to insert. Automatic insertion requires a suitable focused text field and a short enough result; it does not insert directly into a terminal. Otherwise the result appears in the panel for you to review and insert.

Edit selected textCopy Anchor Link

Select the text in another app, use Edit Selection by Voice, and say an instruction such as “Make this shorter, but keep the dates.”

Redcast checks the selection again before replacing it. If the selection changed, the app cannot write it through Accessibility, or the destination is a terminal, the result goes to the preview panel instead. A selection that cannot be read stops the operation before sending audio.

Ask an agentCopy Anchor Link

The Answer with setting can route a voice question to an agent. The speech is transcribed and becomes a background task with the captured window attached. Its result is an ordinary AI Chat conversation, subject to that agent's tools and permissions.

This is useful for an action needing files or a multi-step investigation. For a direct explanation of a paragraph, the ordinary voice-model answer is usually enough.

History and privacyCopy Anchor Link

Ask by Voice and voice edits do not keep their audio. Include Ask by Voice and voice edits in Dictation History is off by default; turning it on stores the resulting text and spoken request. An agent's voice task is already stored as a chat instead.

Captured text and images are sent to the selected route. A password field is excluded from text capture, but the rest of a screenshot can still contain sensitive information. Review Screen Awareness when choosing what to include.

If it does not workCopy Anchor Link

Check AI, Screen Awareness, the approved audio connection, Microphone and Accessibility. A screenshot also needs Screen Recording. If the model connection was disabled or removed, choose an approved replacement; the voice feature deliberately does not choose a different provider by itself.