Gemini CLI, in your own voice
Say the task the way you would explain it to a teammate. DanDin types it into the Gemini CLI prompt, and you send it when you are ready.
Yes, you can talk to Google’s Gemini CLI in Arabic, on Mac and Windows. Put the cursor in the Gemini CLI prompt in your terminal, hold the start key, talk, and let go. DanDin pastes your request into the prompt and never presses Enter. Gemini CLI has an experimental Voice Mode of its own, off unless you turn it on in settings, and the pages we read give no list of its languages.
How to voice type in Gemini CLI
On a Mac
- Open Terminal, iTerm2, Warp or Ghostty, go to your project folder, and run gemini.
- When the Gemini CLI prompt shows up, make sure the cursor is in it.
- Hold Fn and say what you want from Gemini: the file, the change, and what you do not want changed. Then let go.
- Read the request, fix any word if you need to, then press Enter.
On Windows
- Open PowerShell or Windows Terminal in your project folder and run gemini.
- Put the cursor in the Gemini CLI prompt.
- If you want the request in English, turn on Translate to English in Dictation options before you talk.
- Hold Ctrl + Win, explain the task, let go of both keys, and send the request when you are ready.
An example in Gemini CLI
Asking for a login page, with Translate to English on
You say
أبي صفحة تسجيل دخول فيها إيميل وكلمة مرور ولو غلط المستخدم ثلاث مرات قفّل الحساب ربع ساعة
DanDin types
I want a login page with an email and password. If the user enters the wrong credentials three times, lock the account for fifteen minutes.
Gemini CLI’s Voice Mode or DanDin?
Gemini CLI has had an experimental voice mode since version 0.41.0 (5 May 2026). Here is what Google’s pages in the Gemini CLI repo say about it, next to what DanDin does:
| DanDin | Gemini CLI Voice Mode | |
|---|---|---|
| Turning it on | Ready once installed and given its permissions | Experimental and off by default; you turn it on in settings [1] |
| How you start it | Fn on a Mac, Ctrl + Win on Windows, in any app | Hold Space and talk [2] |
| Where your voice goes | To DanDin’s servers, so it needs the internet | By default (Gemini Live) to Google Cloud; a Whisper option runs on your computer [3] |
| Arabic | Arabic and English, even in one sentence | The pages we read give no list of languages [1] |
| Where it works | Gemini CLI, Codex CLI, Claude Code, your editor, and every other app | Inside Gemini CLI only [1] |
If Voice Mode is enough for you and you do not mind turning on an experimental feature yourself, it is right inside Gemini CLI. If you want to explain in Arabic with English mixed in, send the request in English, or have one key across your terminal tools and the rest of your apps, DanDin is built for that.
Where these facts come from
Every cell in the second column is taken from that company’s own page. Here is the text as we read it, the link, and the day we read it.
[1] Gemini CLI repo: the configuration reference, experimental.voiceMode, default false (the word Arabic is not on the page)
Enable experimental voice dictation and commands (/voice, /voice model).
https://github.com/google-gemini/gemini-cli/blob/main/docs/reference/configuration.md (read )
[2] Gemini CLI repo: keyboard shortcuts, app.voiceModePTT on the Space key
Hold to speak in Voice Mode.
https://github.com/google-gemini/gemini-cli/blob/main/docs/reference/keyboard-shortcuts.md (read )
[3] The same page, experimental.voice.backend (default gemini-live) and experimental.voice.whisperModel
The backend to use for voice transcription. Note: When using the Gemini Live backend, voice recordings are sent to Google Cloud for transcription. ... The Whisper model to use for local transcription.
https://github.com/google-gemini/gemini-cli/blob/main/docs/reference/configuration.md (read )
[4] Gemini CLI changelog, version 0.41.0 (5 May 2026)
Real-time Voice Mode: Implemented real-time voice mode with cloud and local backends
https://github.com/google-gemini/gemini-cli/blob/main/docs/changelogs/index.md (read )
Good to know in Gemini CLI
- On a Mac, Secure Keyboard Entry in your terminal blocks every paste. DanDin tells you and the text is not lost, and you turn it off from the terminal’s menu.
- DanDin never presses Enter, so Gemini CLI only works on your request once you send it.
- On a Mac, terminals count as AI tools, so your Arabic speech is written as a clear instruction: one step per line, file names and commands spelled exactly. On Windows it is written tidy as usual.
- If you turn on Gemini CLI’s Voice Mode, use one of the two at a time, so the words are not typed twice.
- It needs the internet, and there is no build for Windows on ARM.
Common questions
Does Gemini CLI have voice input?
Yes, an experimental Voice Mode since version 0.41.0. According to its pages on GitHub (read 19 October 2026), you turn it on with the experimental.voiceMode setting, then hold Space and talk, and the pages give no list of languages. DanDin works in Gemini CLI with nothing to turn on inside Gemini CLI.
How do I talk to Gemini CLI in Arabic?
Run gemini in your terminal and click in its prompt. Then hold the start key (Fn on a Mac, Ctrl + Win on Windows), say the task in your own words, and let go. The request is typed into the prompt, and you send it.
Can I speak Arabic and have Gemini get English?
Yes. Turn on Translate to English in Dictation options on the small DanDin pill, speak Arabic, and the request is typed into Gemini CLI in English.
Does the same key work in Codex CLI and Claude Code?
Yes. DanDin types into any box that has the cursor, so you use the same key with Gemini CLI, Codex CLI and Claude Code, even in the same terminal.