Skip to main content

Voice

You can drive Celeris by talking to it. Voice is a way of working rather than a place to go. Dictation and spoken replies sit in the ordinary chat window, and everything you say runs under the same approvals as anything you type.

Dictation​

Click the microphone in the chat composer, or in the capture overlay, and speak. Your words appear in the input as you say them, and the task fires when a sentence ends. Keep talking and each sentence queues as the next task while you watch the previous one run. When a spoken task finishes, Celeris reads the result back.

Push to talk​

A click leaves the microphone open until you click again, which means it hears whatever the room says in between. To hear only you, hold instead: hold the push-to-talk key while you speak and let it go when you are done. The microphone opens the moment the key goes down, and releasing it sends what you said as one task. Holding the microphone button does the same thing; a quick click still toggles.

The key is the right Option key on a Mac and the right Alt key elsewhere, and you can change it or turn it off in Settings → Voice. Pick a key that types nothing on its own. It works while a Celeris window is in front.

Pressing the key while Celeris is reading a reply cuts the reply short: talking over Celeris is how you interrupt it.

The settings for this live in Settings → Voice.

  • Auto-run each sentence. Turn it off and dictation only fills the input, so you press Enter yourself. With push to talk, letting go of the key fills the input the same way.
  • Push-to-talk key. The key to hold while dictating.
  • Spoken replies. Whether Celeris reads results aloud.
  • Playback speed. How fast replies are read, from 0.5× to 3×.
  • Voice. Which voice reads them, when your speech provider offers a choice.

The player​

While Celeris is reading a reply, a player appears above the composer.

  • Play and pause pick up where the reply left off. Stop silences it and leaves it there to play again.
  • Previous and next step between the answers spoken in this session, so you can hear the last one again without asking again.
  • Speed and voice can be changed while a reply is playing. Whatever you pick becomes your default, and the player says so when it saves.

A change of speed is instant. A change of voice takes effect at the next sentence, because the sentence you are hearing has already been recorded.

Replies start speaking a sentence at a time rather than waiting for the whole answer to be recorded, and each sentence starts playing as soon as its first sound arrives rather than when the last one has. Celeris also begins recording the first sentence while the answer is still being written, so by the time the answer lands there is usually nothing left to wait for.

While the first sentence is being prepared, its segment of the progress bar under the controls fills faintly toward the moment Celeris expects it to start, then fills solidly as it plays. If that moment is a bad one, press Escape or the push-to-talk key and the reply is cancelled before it says a word. Escape stops the task too, if one is still running. It has always meant stop, and cutting the voice does not spare the work behind it.

Voice needs a speech provider key, set in Settings → Voice. Celeris stores it in your operating system's keychain, and it never reaches the app's interface.

The microphone does not stay hot. Dictation switches itself off after a minute of silence, and capture goes quiet while Celeris is speaking, so it never transcribes its own replies.

Push-to-talk​

Push-to-talk lets you hold one key to dictate and release it to send, so a demo microphone hears only you. Set it in Settings → Voice: click the field and press the key you want. Pick a key that types nothing on its own, such as a right-hand modifier or a function key, so holding it never leaves stray characters in whatever window is in front. The clear control turns push-to-talk off again.

Live​

Dictation is for a task at a time, with the screen still in front of you. Live is the conversation that goes on without it: one long exchange that listens until you end it and speaks everything back. It is a Labs feature.

Turning it off​

Voice is opt-in. Leave the provider key unset and Celeris captures no audio at all. Turning spoken replies off in Settings → Voice keeps dictation and stops Celeris reading anything aloud.