---

name: capture-listening description: "Guides SpeechRecognition microphone and streaming capture workflows with Microphone, record, listen, background callbacks, calibration, device listing, thresholds, and output capture." disable-model-invocation: true metadata: disco-role: operating license: NOASSERTION

Capture and Listening

Use this sub-skill when the task is to obtain audio from a microphone or other live AudioSource before transcription: choosing devices, calibrating ambient noise, using Recognizer.record, Recognizer.listen, listen(stream=True), or listen_in_background, handling thresholds, and saving captured audio for later use.

Route here for

• Listing microphones and selecting Microphone(device_index=...).
• Building interactive microphone capture flows with timeout, phrase_time_limit, calibration, and optional WAV output.
• Distinguishing fixed-duration capture with record from speech-boundary capture with listen.
• Streaming chunk capture with listen(stream=True).
• Background capture callbacks, worker queues, and safe stopper patterns.
• Diagnosing PyAudio, default-device, threshold, blocking, ALSA/JACK, and callback-thread issues.

Route elsewhere

• Audio file formats, AudioData conversion methods, splitting, FLAC/WAV/AIFF details, and file-input workflows: use the sibling audio-data sub-skill.
• Recognition engines, language options, remote-service setup, network/model errors, and transcription output: use the sibling recognition-engines sub-skill.
• Installing PyAudio/PortAudio extras, the sprc console script, model downloads, and cross-cutting package setup: use the root skill or sibling cli-model-setup sub-skill.

Read next

• Read references/workflows.md for verified API facts, capture recipes, threshold tuning, streaming/background patterns, and safe output capture.
• Read references/troubleshooting.md when microphone capture fails, hangs, triggers on noise, misses speech, or has background-thread issues.
• Use scripts/microphone_capture_template.py as a safe interactive template. It supports --help without PyAudio, --list-devices, capture time limits, and optional WAV output without recognition-engine calls.

Operating reminders

• Always open Microphone or another AudioSource with a with block before calling record, listen, or adjust_for_ambient_noise; listen_in_background opens the source internally.
• Calibrate with adjust_for_ambient_noise during silence before a first listen when ambient noise is unknown.
• Keep capture separate from transcription. Capture returns AudioData; pass it to an engine only after selecting the correct recognition sub-skill.
• Treat microphone examples as hardware-only native candidates; do not execute them automatically in headless verification.