Speech to Text

If you would rather talk than type, Atua supports two voice workflows:

  • Voice input in the command bar when you want to dictate a prompt before sending it
  • Voice Paste when you want to speak directly into another app and have Atua paste the transcript back for you

Set It Up

  1. Open Settings → General.
  2. In Speech to Text, choose the engine you want to use:
    • Parakeet TDT v3 as the recommended default for fast English dictation
    • Whisper if you want selectable model sizes or broader compatibility
    • Qwen3 ASR for multilingual local transcription
  3. If you choose Whisper, optionally pick a Variant such as Tiny, Base, Small, or Large v3 Turbo.
  4. Choose a microphone input or leave it on System Default.
  5. Click Download Model for the selected engine.
  6. Wait until the status says Ready.

Choose The Right Engine

  • Parakeet TDT v3 is the recommended default when you mainly dictate in English and want quick startup.
  • Whisper runs locally with selectable model sizes. Tiny loads fastest, while larger variants usually improve accuracy.
  • Qwen3 ASR is better when you need broader language coverage, but it is a larger model.

Use Voice Input In The Command Bar

  1. Open the command bar with ⌥ + Space.
  2. Click the microphone button or press ⌘ + M.
  3. Speak normally.
  4. Click the microphone again or press ⌘ + M to stop.
  5. Your words appear in the input field, ready to send.

This mode only fills the command bar input. Atua does not send anything until you press Enter.

Use Voice Paste

Voice Paste is for speaking into another app without opening the command bar.

  1. Put your cursor where you want the text to appear.
  2. Press ⌥ + ⇧ + Space to start recording.
  3. Speak normally.
  4. Press ⌥ + ⇧ + Space again to stop.
  5. Atua transcribes your speech, switches back to the app you were using, and pastes the text there automatically.

While recording, you will see a floating widget near the bottom of the screen. It shows the Atua logo on the left, a live waveform while recording, Listening… during capture, and Transcribing… after you stop.

Use {{transcript}} In A Command

If a saved command includes {{transcript}}, Atua records your voice first and then runs the command with that transcript.

The flow is:

  1. Run the command.
  2. Speak when the recorder appears.
  3. Stop the recording using the same voice control.
  4. Atua inserts the transcript into the command and continues automatically.

This is useful for things like turning spoken notes into an email, summary, or task list.

Good To Know

  • Transcription runs on your Mac after the selected model is installed.
  • The first download for these local speech models comes from Hugging Face.
  • After the model is installed, your voice stays local.
  • You can switch between engines later in Settings → General.
  • Whisper variants trade load time, memory use, and accuracy against each other.
  • Atua keeps the speech runtime loaded for a while after use, so repeated recordings are usually faster than the first load.
  • You can remove the selected model later if you want to free up space.
  • The first time you use it, macOS will ask for microphone permission.
  • Voice Paste saves the transcript and pastes it back into the app you were using before recording started.