← All guides

Warp terminal voice input

Warp already has native voice input. Start there.

Warp’s current Voice feature works in Agent Mode, terminal command input, Find, and its other input editors. On a Mac, the default documented hotkey is hold Fn, speak, then release to transcribe. That is the shortest path when Warp is the destination; the important next step is separating a natural-language agent prompt from exact shell syntax and reviewing either before it can act.

Source-reviewed

1. Enable Warp’s native Voice feature first

Open Warp Settings → Agents → Warp Agent → Voice, turn Voice on, and choose one of the available activation hotkeys. First use requires the operating system’s microphone permission; on a Mac, check System Settings → Privacy & Security → Microphone if Warp cannot hear you.

In Agent Mode, click the microphone button to start and click it again to stop. The current macOS hotkey procedure is hold Fn, speak while holding it, then release Fn to stop recording and transcribe. Because the shortcut is configurable, trust the live setting in the installed Warp build over an old screenshot.

2. Match the voice surface to the job

Warp says Voice is not limited to Agent Mode. It works across its input interfaces, including the terminal command input, Find dialog, and other input editors.

Those destinations do not have the same consequence. A natural-language Agent Mode prompt asks Warp’s agent to interpret a goal. Text in the command input can become exact shell syntax. Find is a search surface. Confirm which input owns focus before recording and verify where the transcript appears.

  • Agent Mode: speak the goal, constraints, and acceptance criteria.
  • Command input: treat every character as potentially executable.
  • Find: use voice for a search phrase, then inspect the results.

3. Keep voice processing separate from later Warp processing

Warp’s Voice documentation names Wispr Flow as the transcription processor. It says voice data is processed in real time and is not retained as a recording after transcription. That is a specific recording-retention statement, not a claim that the complete Warp workflow is local.

After transcription, the resulting text can enter Warp’s command, search, Agent, history, sharing, telemetry, or cloud features. Warp’s privacy documentation says terminal content may be used by enabled cloud and AI features and provides separate telemetry and data controls. Review the active Warp settings, plan, account, and destination rather than transferring the voice-recording statement to every later stage.

4. Review commands and agent prompts before they act

Warp’s own troubleshooting guidance recommends reviewing the transcript before sending, especially for filenames and commands. A misheard flag, path, package, branch, host, negation, or number can change the result.

For exact or consequential shell input, type or verify every character. In Agent Mode, inspect the full prompt and the agent’s proposed actions. Warp documents configurable profiles, approvals, allowlists, denylists, and Full Terminal Use controls; transcription does not replace those execution permissions.

  • Stop voice input before any send or execution action.
  • Confirm the active directory, repository, shell, host, and account.
  • Use explicit approval for sensitive commands instead of treating a transcript as authorization.

5. Troubleshoot the correct layer

If Voice does not activate, confirm that it is enabled, inspect the configured hotkey for conflicts, update Warp, and check the operating system’s microphone permission and selected input device. Restarting Warp can reset a missing microphone according to the current troubleshooting page.

If the microphone works but transcription quality is poor, reduce background noise, move closer to the microphone, check input levels, speak complete phrases, and enunciate filenames or commands. Enterprise administrators can disable Voice or leave it pending approval, so organization policy can be the final boundary.

6. Add IraVoice only for the cross-app or reviewable-brief job

Warp’s native Voice path should be the first choice when all speech belongs in Warp and its Wispr Flow processing boundary fits the task. Installing another tool solely to duplicate that button adds friction.

IraVoice 0.7.1 is a separate option for macOS 26+ Apple-silicon Macs when speech recognition and supported formatting should run on the Mac across supported developer, browser, email, and document fields. Voice-to-Spec can create reviewable Markdown before you choose Warp or another agent. IraVoice does not integrate with Warp, read its terminal, press Enter, approve commands, control Agent Mode, or bypass Warp permissions.

Decision guide

Which Warp voice path fits the task?

Choose by destination, processing boundary, and whether exact executable syntax is involved.

If you needChooseWhy
Speak directly into Warp Agent ModeWarp Voice microphone or hotkeyIt is the native path and understands the active Warp input surface.
Search inside Warp by voiceWarp Voice in FindWarp documents Voice across its Find and other input editors.
Enter an exact shell commandType and verify the syntaxOne transcription error can change execution.
Keep speech recognition on a compatible MacIraVoiceWarp documents Wispr Flow processing; IraVoice documents on-Mac recognition.
Create reusable Markdown before choosing an agentIraVoice Voice-to-SpecThe brief stays reviewable and is manually pasted into Warp.

Primary sources

Use the current documentation.

This guide was reviewed July 26, 2026. Product versions, settings, policies, and availability can change.

  1. Warp: Voice input for agents

    Current settings path, microphone button, configurable hotkeys, Mac Fn workflow, input scope, Wispr Flow processing, non-retention statement, limits, and troubleshooting.

  2. Warp: Privacy overview

    Current offline terminal boundary, cloud and AI features, terminal context, telemetry controls, login, stored items, and enterprise retention options.

  3. Warp: Full Terminal Use

    Current live terminal access, proposed commands, approvals, takeover controls, and profile-level autonomy.

  4. Warp: Agent profiles and permissions

    Current action permissions, autonomy levels, command allowlist and denylist, MCP controls, and run-until-completion boundary.

Questions

What readers usually ask next.

Does Warp terminal have built-in voice input?

Yes. Warp documents native Voice for Agent Mode and its other input interfaces, including terminal command input, Find, and additional editors.

What is the Warp voice-input shortcut on Mac?

The current documentation says to hold Fn while speaking and release it to stop and transcribe. The shortcut is configurable, so verify the current choice under Settings → Agents → Warp Agent → Voice.

Is Warp voice input processed locally?

Warp’s current Voice page does not describe a local-only path. It says Wispr Flow powers transcription, processes voice data in real time, and does not retain it as a recording after transcription.

Does Warp voice input execute a command automatically?

The Voice page documents recording and transcription across Warp inputs, not blanket execution authorization. Review the resulting text and keep command execution or Agent approvals as separate decisions.

Can IraVoice control Warp Agent Mode?

No. IraVoice has no Warp integration and cannot read the terminal, approve actions, press Enter, or change Warp permissions. It can produce local text or reviewable Markdown that you manually place in the intended Warp field.

macOS 26+ · Apple-silicon

Try the local Mac workflow for yourself.

Dictate ordinary text or turn a rough technical thought into reviewable Markdown during the 14-day full trial.