A new local command-line tool called erm removes speech disfluencies from English audio recordings by combining faster-whisper transcription, direct audio analysis, and ffmpeg. The release highlights the surprising technical complexities involved in cleanly identifying and excising filler words from spoken audio.