Gemini CLI voice input is not listed in the official command reference as of 5 November 2026, so by default you type prompts. You can add voice with a community extension on GitHub, or dictate with Windows voice typing, Mac Dictation or Genie 007, which types wherever your cursor is.
Gemini CLI Voice Input: What Exists Today
Gemini CLI is Google’s terminal tool for talking to Gemini models. Its docs describe it this way: “Gemini CLI brings the power of Gemini models directly into your terminal.” You type a prompt, press Return and read the answer in the same window.
That leaves three practical routes if you would rather speak. You can install a community voice extension, you can use the dictation tool built into Windows or macOS, or you can use a dictation app that types at the cursor. All three end the same way: words appear in the prompt line and you press Return yourself.
The same ideas apply to voice dictation for Claude Code and Codex dictation.
Does Gemini CLI Have a Built-In Voice Mode or /voice Command?
On 5 November 2026 the official command reference lists no voice command. It names 38 slash commands, from /about and /agents to /upgrade and /vim, plus @ commands for files and ! commands for the shell. None of it mentions voice, audio, dictation or a microphone.
That is not the whole picture, so here is the honest version. A pull request titled “feat(voice): implement real-time voice mode with cloud and local backends” was merged into the google-gemini/gemini-cli repository on 24 April 2026. Its description says it lets users “dictate prompts directly into the terminal” and it adds a /voice slash command. The page also shows a later commit that placed voice mode behind an experimental flag.
The command reference does not show that this experimental mode ships in a current release, so do not count on a built-in Gemini CLI voice mode being there. Check /help in your own install: if /voice is not listed, it is not available to you.
The Community Voice Extension: Install, Auth, Limits
There is a community extension on GitHub (kstonekuan) that adds a /voice command to Gemini CLI. Its README opens with “Voice mode for Gemini CLI. Talk to Gemini from your terminal, powered by the Gemini Live API.” It is not a Google product.
It needs three things: Gemini CLI, Node.js v18 or newer and a Gemini API key. The README gives two install commands. Either one works, then you save your key.
| Step | Command or action |
|---|---|
| Install from GitHub | gemini extensions install https://github.com/kstonekuan/gemini-cli-voice-extension |
| Or install from npm | gemini extensions install @kstonekuan/gemini-voice |
| Save your API key | gemini-voice auth |
| Start listening | Type /voice inside Gemini CLI |
The README is open about its limits. There is no push-to-talk: you type /voice to start listening, and the README calls that a platform limitation. Gemini CLI does not support live output from the extension’s subprocess, so the interactive display is suppressed. It also says the model repeats back what it heard before it answers.
The author describes the project as “a stepping stone” towards a native voice mode with push-to-talk. Treat it as an extra to try, and install extensions only from sources you trust.
Gemini CLI Voice to Text With System Dictation (Windows + H, Mac Fn)
A terminal prompt is a text input, so the operating system’s own dictation can fill it. This needs no Gemini setup, and it is the lowest-risk first step for Gemini CLI dictation. Gemini CLI voice to text this way is plain transcription: you get your words, nothing more.
| Windows 11 voice typing | Mac Dictation | |
|---|---|---|
| Start | Press Windows key + H on a hardware keyboard | Press and release the Microphone key, use your Dictation shortcut, or choose Edit > Start Dictation |
| Needs | A working microphone, the cursor in a text box and an internet connection | Dictation turned on in Keyboard settings |
| Stop | Say “Stop listening” or press the microphone button | Press Escape, the Microphone key or the shortcut again |
| Limit | Settings has a Default microphone option in the flyout’s gear menu | Stops after 30 seconds of silence |
These details come from Microsoft’s “Use voice typing” page and Apple’s Dictation guide. Apple adds that you can dictate text of any length without a timeout, so the 30 seconds applies to silence only.
Whether a given terminal window accepts system dictation can vary, so click into the prompt and try a short sentence first. For the platform basics see our Windows and Mac pages.
Gemini CLI Voice Input Not Working? Checks in Order
Most failures come from focus, the microphone or a missing setup step, so work through these from the easiest to the rarest.
- Check that your operating system uses the right microphone as its default input, and that it works in another app.
- Click inside the terminal so the cursor sits in the Gemini CLI prompt before you start Windows + H or Dictation.
- For the extension, confirm it is installed with gemini extensions install and that you ran gemini-voice auth to save your API key.
- Type /voice to start listening, because the README says there is no push-to-talk and no hotkey to hold and talk.
- Check Node.js is v18 or newer, and that you are online, since Windows voice typing needs a connection and the extension relies on the Gemini Live API.
If the standalone tool from the same README is installed, gemini-voice devices lists your audio input devices and gemini-voice transcribe –device 1 picks a specific one. That helps when the wrong microphone is in use.
A missing /voice command in Gemini CLI is the expected result unless you installed the extension. Our Antigravity voice input guide follows the same logic for another Google tool.
Gemini CLI vs Claude Code, Codex and Copilot CLI: Who Has Native Voice
Voice support differs between AI coding tools and changes often. Each tool has its own guide stating what the maker’s pages said on the day it was written.
For Anthropic’s tool, read voice dictation for Claude Code. For OpenAI’s, read Codex dictation. For Google’s editor, read Antigravity voice input. Our other developer guides sit in the voice typing guides hub.
Dictate Prompts to Gemini CLI With Genie 007 on Windows and Mac
Genie 007 is our app, and it types where your cursor is in any app, including a terminal, on Windows and Mac. So you click into the Gemini CLI prompt, speak, and the text lands in the line. There is no extension to install, no API key to paste and no /voice command to remember.
The limit of plain transcription shows up fast in a terminal. A literal transcript of a rambling spoken thought makes a poor prompt, and you end up editing it by hand. Genie 007 is built around think-to-text: it understands the context and delivers what you mean, not just the words you said.
A concrete example: you say, “ask it to find why the login test fails and suggest a fix, keep it short.” The prompt that arrives is a tidy request, ready to send.
Terminal Dictation Tips: Commands, Paths and Code Words
Terminals are less forgiving than chat boxes, so a few habits help. Dictate the plain-English part of the prompt and type the exact bits yourself.
- Say the task out loud, then type file paths and commands such as @src/app.js, because a spelled-out path is slow to speak and easy to mishear.
- Read the line before you press Return, because dictation can mishear function and library names.
- Keep each dictated prompt to one job, which makes the result easier to review.
Never dictate a command you would not run on your own.
Frequently Asked Questions
Does Gemini CLI have voice input?
On 5 November 2026 the official Gemini CLI command reference lists no voice command, so nothing is documented there. A merged pull request added an experimental /voice mode in April 2026, but check /help in your own install. A community extension on GitHub (kstonekuan) also adds a /voice command.
How do I add a /voice command to Gemini CLI?
Install the community extension from GitHub (kstonekuan) with gemini extensions install https://github.com/kstonekuan/gemini-cli-voice-extension, then run gemini-voice auth to save your Gemini API key. It needs Node.js v18 or newer. After that, type /voice inside Gemini CLI to start listening.
Can I use Windows voice typing or Mac Dictation in the Gemini CLI terminal?
Often yes, because the prompt is a text input. On Windows press Windows key + H with your cursor in the prompt. On a Mac press the Microphone key or your Dictation shortcut. Click into the terminal first and test a short sentence, as terminals can vary.
Does Gemini CLI voice work offline?
The extension’s README says it is powered by the Gemini Live API and needs a Gemini API key, so it is a cloud service rather than an offline one. The README does not describe an offline mode. Windows voice typing also needs an internet connection, according to Microsoft.
Try Genie 007 Free
If you want to speak to Gemini CLI without an extension or an API key, Genie 007 types your prompt at the cursor and turns a spoken thought into a clear request.
Request access to the desktop app and try it in your own terminal. Genie 007 works on Windows, Mac, mobile and as a browser extension.
Written by Bill Kiani, founder of Genie 007.
