--- title: "Voice: dictate, read aloud and talk to your Mac" description: "How do I use voice in ToolPiper: dictation, reading text aloud, and voice chat or voice commands?" date: 2026-09-05 author: "ModelPiper" tags: ["Voice", "Dictation", "Speech to Text", "Text to Speech", "Hotkeys"] type: "faq" canonical: "https://modelpiper.com/help/voice-dictation-and-commands" --- # Voice: dictate, read aloud and talk to your Mac > How do I use voice in ToolPiper: dictation, reading text aloud, and voice chat or voice commands? Voice in ToolPiper lives on two keys that work in every app on your Mac. Hold the right Option key and speak to dictate into whatever has the cursor. Tap the right Option key to hear the text you have selected read aloud. Hold the right Command key and speak to talk to the assistant, which can answer out loud or carry out a command on your Mac. Everything runs on this Mac with local speech models, so it works offline, nothing is uploaded, and none of it is metered or paid. ## Set up in the Home pane Open ToolPiper's Home pane and use **Try dictation** and **Try voice command**. Each one asks for what it needs the first time: the microphone, the Accessibility permission that lets the keys work system-wide and paste the transcript, and the speech-to-text model download. Say something while it shows **Listening** and it confirms with **Done, it works**. The first use after ToolPiper launches can take several seconds while the speech model warms up. After that a transcript arrives in well under a second. ## Dictate: hold right Option Hold the right Option key, speak, and release. The transcript is pasted where your cursor is, in any app. Hold the key for at least a second; a shorter press is treated as a tap, which reads your selection aloud instead, or shows a hint to hold longer. The default English model returns lowercase text without punctuation. To get capitals and punctuation, pick the **Parakeet Unified** preset in the Speech-to-Text row of the Home pane. For languages other than English, pick **Parakeet TDT V3**, which covers twenty-five languages and Japanese. Back-to-back dictations are joined with a space automatically. ## Read aloud: tap right Option Select any text, tap the right Option key, and it is read aloud. Tap again to stop. This works on day one with the built-in system voice and switches to ToolPiper's own higher-quality voices once a text-to-speech model is downloaded. Text you select in a password field is never read, and a banner tells you why. The same command is available from the Services menu as **Speak with ToolPiper**. ## Talk to your Mac: hold right Command Hold the right Command key, speak, and release to send. Your words go through speech-to-text, then to the model you have loaded, then back to you. The model can also use ToolPiper's tools to act on your Mac: change the volume or brightness, arrange windows, toggle focus modes, control media playback, and more. Anything risky asks for approval first through a macOS notification. One switch decides how the assistant answers, under **Settings > Voice > Voice Chat**: - **Conversation on** (the default): the assistant speaks its reply, in the Cosette voice unless you pick another. - **Conversation off**: voice command mode. Nothing is spoken. The assistant carries out what you said and posts one notification naming what it did, for example "Ran system volume set", or asking you to say it again if it did not catch a command. **Full-Duplex** in the same settings makes it hands-free: no key to hold, the assistant listens continuously and you can interrupt it while it is speaking. ## Which model answers Voice chat uses whatever text model is loaded in ToolPiper. You can point it at a specific local model or a cloud provider you have connected using the provider picker in **Settings > Voice > Voice Chat**. A voice command needs a loaded text model too; if none is loaded, load one first. ## Use your own voice Under **Settings > Voice**, set a reference clip of yourself speaking and turn on **Use my voice everywhere**. The first ten seconds of the clip are used, and a clip under one second is refused. Once it is on, every spoken reply and every read-aloud uses that voice until you switch it off or remove the clip. ## Where voice is not The keys belong to ToolPiper, so voice works from any application on your Mac but not inside the web app in a browser. If a key does nothing, check that ToolPiper is running and that the Accessibility permission is granted in **System Settings > Privacy & Security > Accessibility**.