Most mammals produce sound via vocal fold vibration, in accordance with the myoelastic-aerodynamic theory. However, when lab mice produce ultrasonic vocalizations (USVs), they do not use vocal fold vibration, but instead produce self-sustained whistles where a glottal air jet impinges on an intralaryngeal structure. The tones are produced by a feedback loop between flow structures traveling downstream and acoustic waves traveling upstream in the flow. We examined sound production in Alston’s singing mouse, Scotinomys teguina. This murid rodent produces a highly elaborate, stereotyped song consisting of rapidly-repeated, frequency-modulated notes that span from 43 to 10 kHz. We analyzed sound production in excised larynges that combined airflow modulation with sound recording and high-speed imaging. In a series of experimental manipulations, we identified the anatomical structures essential to sound production and manipulated laryngeal cartilages to imitate muscle action and identify how frequency is modulated. Along with in vitro manipulations, we used histology and μCT for morphometric analysis. We found that singing mice sing by producing a jet that impinges on the alar cartilage. However, unlike lab mice, singing mice require the inflation of a hypertrophied ventral pouch. In vitro manipulations show that frequency is set predominantly by jet speed, which can be modulated by changes in air flow and glottal area. Thus, the fine interplay of action of respiratory and laryngeal muscles, such as the cricothyroid, and thyroarytenoid, combine to control frequency.