End Transcribing

Ends the microphone opened by Start Transcribing and turns what was said into text, with an AI model or on the phone itself.

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

Declared ports

Port	Label	Flow	Type
in	in	execution in	—
out	out	execution out	—
answer	Answer	data out	String
language	Language	data out	String

Configuration

Field	Key	Default	Notes
If it fails	fallback	—	—

Needs

• 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.

Notes

End Transcribing ends the microphone opened by Start Transcribing, sends what it heard to an AI model, and puts the answer on the answer port. Leave the question empty and you get a transcript.

• It has one field, If it fails. Everything else was decided on Start Transcribing.
• It also collects a listen that already ended on its own limit.
• The sound is discarded once it has been sent. There is nothing to clean up.

Working of the Node

• The node ends the running listen and waits for the last fragment of sound, so nothing is cut off.
• A listen that already hit its limit is collected here instead. The sound was kept for exactly this.
• With What to ask empty the sound goes for transcription; with it filled in the sound and the question go to a chat model together.
• The reply lands on answer. Every failure lands on If it fails and the node still pulses out.

Example: a hands-free note

• Add Start Transcribing with Listen for at most set to 60.
• Wire out into whatever should happen while it listens.
• Wire that into End Transcribing. Pick a Model and leave What to ask empty.
• Wire answer into Write File.

Explanation:

• The empty question makes this a transcript rather than a summary.
• The 60-second limit means the file still gets written even if the macro takes a detour and reaches this node late.

When Nothing Was Started

• Reaching this node with no listen running puts If it fails on the port and writes an error to the console.
• That is reported rather than ignored, because a macro reaching a stop it never started is usually a graph running in an order its author did not expect.
• A listen that heard only silence is different. That writes one ordinary line to the console and is not an error.

Points to Remember

• The language port carries the code the recogniser says it used, and is empty when nothing reported one — every AI transcription, and every phone older than Android 14.
• This node makes no choices at all. Start Transcribing picked the engine and the model, and this one collects whichever session is open.
• An empty question is a setting. It routes the sound to the service built for transcripts, which is cheaper and works on servers no chat model runs on.
• Claude cannot hear. Point Model at Gemini, OpenAI, OpenRouter or a self-hosted model.
• The reply is cut off at Longest reply (tokens), reaches the port anyway, and says so in the console.
• Both nodes need the microphone permission, even though only the start node opens it.