audio-injection-testing

Audio injection testing automates wake word detection by injecting pre-recorded audio into virtual audio pathways instead of requiring manual speech. This enables running hundreds of test cycles on Linux and Windows to validate wake word detection accuracy and consistency without repetition.

Yes. This Skill injects pre-recorded audio files into virtual devices and captures transcription results, allowing you to compare STT output against known phrases and measure accuracy across test sequences with configurable delays and logging.

Audio injection testing requires running virtual audio setup scripts, configuring devices by index or name, and preparing pre-recorded test audio files. The Skill provides setup automation, device configuration, test file sequencing, and cleanup procedures for repeatable, auditable testing across both platforms.

Execute test sequences injecting audio through the full pipeline: wake word detection triggers STT processing, which feeds the LLM, and TTS generates responses. Logging captures each stage, enabling end-to-end validation of wake word, transcription, and response generation in a single automated test run.

The Skill accepts pre-recorded audio files for injection testing. Supported formats include MP3 and other standard audio formats compatible with virtual audio device input on Linux and Windows, enabling flexible test asset management and sequencing.

Audio injection eliminates variability from ambient noise, microphone hardware, and repeated manual speech, enabling deterministic, reproducible test results. Virtual audio pathways ensure consistent input delivery, comprehensive logging, and scaled test execution without environmental interference.