Our daily life abounds in various characteristic voices, such as train announcements, vendor's cries, and anime dialog. Although phonetic science in general has not paid much attention to these voices, each of them seems to have its own phonation, resonance, phrasing, and intonation. Can we capture these voices “in residue”? If so, how, and to what degree? As a case study for constructing prerequisite for examining such questions, this paper deals with Japanese idioms kuchi-o togaraseru (purse one's lips) and kuchi-o yugameru (curl one's lips) and their surroundings, and clarifies how they work in daily communication by using audio-visual data.