To dictate a prompt into IntelliJ IDEA on Mac, open the AI Chat tool window, click the chat input, dictate with your Speakmac shortcut, and read the text before pressing Enter. If the text does not land in the field, use Copy Last Transcript in Speakmac and paste it. Apple Dictation is worth one test, but JetBrains' own issue tracker has open reports of it duplicating or dropping text in AI chat fields.

IntelliJ IDEA is JetBrains' IDE; the same steps apply to PyCharm, WebStorm, GoLand and the other IntelliJ-based IDEs. The prompt field belongs to JetBrains AI Assistant, which hosts plain Chat plus agents: Junie (JetBrains' own coding agent), Claude Agent, Codex, and any agent you add through the Agent Client Protocol. The newer Air in IDEs plugin is a separate surface, covered at the end.

Where does the prompt go in a JetBrains IDE?

Click AI Chat on the right toolbar. The input sits at the bottom of the tool window, with the chat mode selector, model selector, Add attachment and Send around it. JetBrains documents that the window opens with an agent selected by default, so check the selector before you start: Chat answers and suggests, while an agent such as Junie can edit files and run commands.

Enter sends; Shift+Enter adds a new line. That matters for voice, because a dictated paragraph should stay in the field until you have read it.

Does Apple Dictation work in JetBrains AI Chat?

Sometimes, but it is the least reliable route. JetBrains' YouTrack has several open reports:

  • LLM-27033 (Open, April 2026): Mac Dictation into the Junie chat produced repeated fragments such as "CCan yCan you Can you ta…" in IntelliJ IDEA 2026.1.
  • JUNIE-122 (Backlog, 21 votes): Dictation not recognised in the Junie input. One commenter with wrist injuries dictates into Notes and pastes instead.
  • IJPL-175022 (Open): Voice Control shows "dictation unavailable" in the editor after macOS 15.2. The listed workaround is to switch Voice Control off and on again after the IDE starts.

If Apple Dictation works in your version, use it. If it duplicates text, stop rather than editing around it; the paste route below is cleaner.

How do you dictate a prompt into AI Chat?

  1. Install Speakmac, grant microphone and Accessibility access, and choose a dictation model.
  2. Open AI Chat and pick Chat, Junie or another agent. For Junie, choose Default or Plan in the mode picker.
  3. Click the chat input so the caret is there. Use your Speakmac Push to Talk or Hands-Free shortcut.
  4. Say the change, the current behaviour, the constraints and the check that proves success. Stop dictation.
  5. Type @ and attach the exact files, symbols or local changes, correct any identifiers, then press Enter.

Speakmac's shortcut lives in Speakmac, not the IDE keymap. Avoid a binding that JetBrains already uses, such as double Control; see shortcut help.

What is a useful voice prompt?

Speak the reasoning, then attach the evidence. For a bug:

CSV exports drop the last row when the result is an exact multiple of the page size. Find where the pagination loop stops, fix it without changing the public method signatures, add a test with exactly two full pages, and run the existing export tests.

Then type @ and attach the export service and its test file.

For something larger, ask Junie in Plan mode to stop before editing:

Plan only. I want to replace the hand-written retry loop in the payment client with our shared backoff helper. List the files you would change, any behaviour that changes, and the tests you would add. Wait for my approval.

These are example briefs, not claims that an agent completed them. Review the diff and run the checks yourself. Instructions you repeat every time belong in an AGENTS.md file for Junie or in AI Assistant project rules, not in every prompt.

What should you type instead?

Speak the explanationType or attach the exact evidence
Expected versus actual behaviourStack traces and test output
Constraints and what must not change@ mentions for files, symbols and commits
Why one approach beats anotherClass names, paths and commands

Do not dictate credentials or private customer data into a prompt. Speakmac transcribes on your Mac, but after you press Enter the selected model provider's policy applies. Speakmac's privacy page covers its own step only.

What if text does not appear?

Click back into the chat input and try one short sentence. If Speakmac shows a transcript but nothing arrived, use Copy Last Transcript and paste with Cmd+V, or assign Paste Last Transcript to a shortcut. Both are described in History and recovery.

If no transcript was produced, check the microphone, model and Accessibility permission in the Help Center. That is a different failure from the IDE dropping input. We have not run a microphone-to-field test in every JetBrains IDE, so treat paste as the dependable route.

Does JetBrains have its own voice input?

Partly. AI Assistant's AI Chat has no documented microphone button; the speech-to-text request JUNIE-796 is still open. Two newer JetBrains surfaces do:

  • Air in IDEs, an Early Access plugin for IntelliJ-based IDEs 2026.2 and later, has a Start Voice Input button in its prompt field.
  • Junie CLI has a /voice command that records from your microphone in the terminal.

If you live in one of those, try the built-in option first. A system-wide app helps when you also want to dictate into the AI Assistant chat, commit messages, review comments, Slack and email, with the same shortcut, your custom words, and on-device transcription.

Try Speakmac for Mac, or compare the workflow with voice dictation for VS Code and voice prompting in Codex.