Voice Dictation on Apple TV

What the dictation bar is

vmuxTV has a dedicated input strip that slides up from the bottom of the screen called the dictation bar. It is the primary way to send text to a running terminal session because the Apple TV does not have a hardware keyboard.

To open the dictation bar, press the Play/Pause button on the Siri Remote while a session is in focus, or press Start / Menu / Options on a paired game controller. Press the same button again to close it.

The bar contains three controls:

Pressing return on the on-screen keyboard while the dictation bar's text field is focused has the same effect as pressing Send.

Command Mode versus Prose Mode

The mode toggle is the most important control on the bar. It changes what happens when you press Send.

Command Mode (default)

In Command Mode, vmuxTV runs your input through a voice command parser before sending it. The parser knows how to translate spoken or typed phrases into terminal keystrokes — including special keys like Tab, Enter, Escape, the arrow keys, function keys, and Ctrl/Alt/Shift modifiers — and it knows the NATO phonetic alphabet ("alpha bravo charlie") so dictation can spell things out reliably.

Examples that work in Command Mode:

If the parser does not recognize a phrase as a command, it falls back to sending the raw text. Command Mode is therefore safe to use for arbitrary input — at worst it sends what you typed.

Prose Mode

Switch to Prose Mode (the text-cursor icon) when you are sending text that happens to contain words like "enter", "escape", or "control" and you want them to be typed literally instead of interpreted.

Typical Prose Mode use cases:

• Typing into a git commit -m "..." message.
• Pasting a paragraph into a remote chat tool.
• Filling out a form inside a vim or nano buffer.

In Prose Mode, the input is sent verbatim — no parsing, no Enter key appended.

Voice dictation

To dictate instead of type:

1. Open the dictation bar with Play/Pause.
2. The text field receives focus and the on-screen keyboard appears.
3. Press the microphone button at the top-right of the on-screen keyboard. Apple TV starts recording.
4. Speak. Words appear in the text field as Apple's speech recognizer transcribes them.
5. Press Done on the keyboard, or wait for the recognizer to finish.
6. Press Send to submit.

vmuxTV requests two privacy permissions on first use of dictation:

• Microphone access — required for capture.
• Speech recognition — required for on-device or server-side transcription.

If you decline either, the microphone button is disabled and you fall back to typing or the iPhone Keyboard. To re-enable: Settings → Apps → vmux → Microphone / Speech Recognition on the Apple TV.

Accuracy tips

Voice recognition on Apple TV is the same engine used in the rest of tvOS. To get cleaner transcripts:

• Speak in short phrases. "list dash l" is more reliable than a long compound command.
• Use NATO when spelling. "alpha bravo charlie" produces abc. "a b c" usually does too, but the NATO alphabet eliminates ambiguity.
• Pause before special keys. Saying "ls dash l ... enter" with a small pause helps the parser identify "enter" as a key rather than a literal word.
• Switch to Prose Mode for natural language. If you are not trying to send keystrokes, Prose Mode skips the parser and avoids surprises.
• Check the field before pressing Send. The transcript is shown in the text field. If it is wrong, edit it with the on-screen keyboard or the iPhone Keyboard before submitting.

What voice cannot do

• It cannot intercept keystrokes mid-typing — the dictation bar is one-shot per Send.
• It cannot send long-running streams (you cannot dictate a continuous monitoring session).
• It cannot reliably dictate code with punctuation. Typing or pasting from the iPhone Keyboard is more reliable for code.
• It cannot dictate passwords. Always type passwords with the iPhone Keyboard or a paired Bluetooth keyboard.

Fallbacks for fast input

Three faster alternatives to dictation:

iPhone Keyboard

When the dictation bar is open, every iPhone or iPad on the same iCloud account that is within Bluetooth range gets a banner notification reading "Apple TV Keyboard". Tap the banner. The phone shows a full keyboard. Type into it; the text appears on the TV in real time. Tap Done on the phone to commit, then press Send on the TV.

This is the fastest way to enter long commands, hostnames, and complex pipelines.

Bluetooth keyboard

If you pair a Bluetooth keyboard through Settings → Remotes and Devices → Bluetooth, every keystroke is forwarded directly to the focused terminal — no dictation bar required. Modifier keys, arrow keys, function keys, Esc, and Tab all work. This is the closest experience to vmuxMac on Apple TV.

When a Bluetooth keyboard is connected, you can leave the dictation bar closed and type directly into the running session.

Game controller

A paired game controller maps its buttons to terminal keystrokes through the same keybind system used on the rest of the vmux family. See Input — Remote and Gamepad for the default mapping. The controller is fast for navigation and a small set of common keys (Enter, Esc, Tab, arrows, Page Up/Down, Ctrl-C, Ctrl-D), but cannot type arbitrary text.

What happens after Send

When you press Send:

1. The text field is read.
2. In Command Mode, the parser converts it into a sequence of key events.
3. In Prose Mode, the text is forwarded verbatim.
4. The events flow into the terminal session over the same path as a hardware keyboard would use, including local echo (your input shows up immediately in the terminal even before the remote acknowledges it).
5. The text field is cleared.
6. The dictation bar stays open so you can send another command without re-opening it.

The dictation bar does not auto-close. Press Play/Pause again, or swipe the focus out of the bar with the clickpad, to dismiss it.

Related

• Input — Remote and Gamepad — full button reference.
• Connecting — getting a session open in the first place.
• Voice Commands — the parsing rules used in Command Mode.
• Troubleshooting — for cases where dictation refuses to start.