A disconnected microphone should not leave a voice app pretending to listen. Detect the lost capability through the supported audio path, stop or settle the affected session, and explain what text or audio remains recoverable.
Changing to another available input is a product decision with privacy and quality implications. Do not silently record from an unexpected device merely because it exists.
Track device and session identity
Record the selected input reference, current session, capture state, and any established transcript segments. Review Apple's current capture-device documentation and the actual audio API used by the application.
Separate device unavailability from permission denial, recognition failure, and silence. These conditions can look similar to a user but require different recovery actions.
Disconnect a fictional external microphone
Start a harmless dictation using a controlled external input, then disconnect it before speech, during speech, and while recognition is finishing. Test reconnecting the same device and selecting another input.
Verify capture stops or enters the defined interrupted state, visible indicators agree, and incomplete output is not inserted as an unquestioned success. Preserve confirmed segments or source audio only under the explicit retention and recovery policy.
Start another session after recovery and check that old buffers, device references, and late recognition results cannot enter it. Include repeated disconnects and a change while the app is not frontmost.
Make the next action explicit
Offer an appropriate device choice, retry, review, or cancellation route. If the app cannot know whether the final sentence was captured, say so rather than inventing a complete transcript.
Keep diagnostics limited to safe stage, device category or necessary reference, version, and failure information. Audio and dictated text should not enter ordinary logs merely to explain a hardware event.
The tradeoff is continuity versus predictable capture. Automatic switching may reduce interruption in some workflows but can surprise users and change sound quality. Requiring confirmation can be safer when the input source materially affects trust.
Use named hardware and macOS versions for verification. No successful recovery, device compatibility, or transcription-quality measurement is claimed here.
The Mac dictation guide explains capture-to-insertion lifecycle, and StackSpeech's portfolio entry provides product context. This disconnect matrix is proposed guidance and does not describe a reported incident there.
Questions about microphone disconnection
Should the app immediately use the built-in microphone?
Only under an explicit understandable policy and supported implementation. A device change should not create unexpected capture.
Can partial text be inserted automatically?
Define the contract carefully. An interrupted session may need review because the app cannot establish complete speech capture.
What should be tested first?
Disconnect during speech, reconnect, and start a new session while delaying the old result. Bring mixed output to a Mac audio-recovery review.