sys2txt is a command-line tool for recording system audio and turning it into text locally. It captures PulseAudio or PipeWire monitor sources with ffmpeg, then uses a supported Whisper engine to produce the transcript. Its documented prerequisites are a modern Linux distribution, ffmpeg and Python 3.10 or later, making it a technical utility rather than a browser transcription service.
The tool supports two recording patterns. An on-demand session transcribes after recording stops, while a live-ish mode divides the recording into time segments and processes them as they are created. The latter can print results continuously, but it is segmented transcription rather than a promise of instantaneous captions. Engine choices include faster-whisper, the reference openai-whisper implementation and whisper.cpp, with selection depending on what is installed and the available hardware.
Output formats include plain text, SubRip subtitles, WebVTT, JSON and tab-separated timing data. In live mode, timed output is rebased onto the full recording timeline so segment cues continue forward. Local processing avoids a hosted transcription API, but users still need to install the software and models and check recognition errors. Recording system output can capture other people's speech or sensitive material, so consent and the rules of the audio source remain the user's responsibility.
