Skip to content
Download

Start Transcribing

Opens the microphone and keeps listening until End Transcribing turns what was said into text.

An action in the AI group. Its type is action.transcribe_start.

Port Label Flow Type
in in execution in
out out execution out
prompt What to ask data in String
Field Key Default Notes
Transcribe with using ai ai, phone
Model modelRef chosen from AI_MODEL; shown when using is ai
What to ask prompt leave empty to transcribe it; shown when using is ai
Longest reply maxOutputTokens 1024 tokens; shown when using is ai
Language languageMode phone phone, detect, chosen; shown when using is phone
Which language language chosen from RECOGNITION_LANGUAGE; shown when languageMode is chosen
Listen for at most maxSeconds 60 seconds
Stop after this much silence silenceSeconds 0 seconds, 0 = listen the whole time
  • microphone access

A missing permission is flagged in the editor rather than blocking the macro — the graph is fine, the phone is not, and granting it starts the node working with no edit here at all.

Start Transcribing opens the microphone and returns straight away, so the macro can carry on while it listens. Nothing is sent anywhere until End Transcribing collects the sound.

  • It carries no Model and no question. Those live on the stop node, where the answer comes out.
  • Listen for at most (seconds) is a limit, not a length. It is required.
  • Stop after this much silence defaults to 0 here, meaning do not stop early.
  • Language offers the phone’s language, automatic detection, or one you choose — the same three answers Transcribe gives, and it is set here because this node opens the session.
  • The node claims the microphone. If a recording or another listen is already running it reports that and changes nothing, because there is one microphone.
  • Reading runs in the background. The node pulses out immediately, whether it started or not.
  • The sound accumulates in memory. Nothing is written to storage at any point.
  • When the limit runs out the microphone is released and the sound is kept, so a stop node reached later still gets what was recorded.
  • Stopping the macro releases the microphone.
  • Add Start Transcribing. Set Listen for at most to 30.
  • Wire out into an Ask Yes or No dialog reading Recording. Done?
  • Wire the dialog’s confirmed port into End Transcribing.
  • Pick a Model on that node, leave What to ask empty, wire answer into a notification.

Explanation:

  • The dialog can only be shown because this node returns instead of blocking. Listen with AI would have held the run for the full 30 seconds first.
  • The 30-second limit is a backstop. If nobody taps the dialog, the microphone still closes.
  • The whole pair is configured on this node: the engine, the model, the question, the language and the two limits. End Transcribing only collects the result.
  • That is the same shape as Record Audio’s family, where Start Recording holds every field and Stop Recording holds none.
  • Model, What to ask and Longest reply appear only when Transcribe with is set to an AI model. On the phone engine they are not shown, and nothing asks you for them.
  • Language appears only on the phone engine, because an AI model works the language out for itself.
  • Zero silence means listen the whole time. That is the opposite of the same field on Listen, which hands the decision to the phone.
  • The phone’s recogniser hears one utterance at a time, so a long session is many short ones joined up. You will not see that happen; it is why a pause does not end the session.
  • A very long pause with nothing said at all can still end it on some phones. The AI engine records straight through regardless.
  • A second Start while one is running reports that and leaves the first alone.
  • Two minutes is the ceiling however large a number you type.
  • The system’s recording indicator stays lit for the whole span. That is Android, not the app.
  • Use Transcribe instead when the macro has nothing to do while it listens. One node is simpler than two.