This paper reports on a system created to help language students learn the vowel inventory of Swedish. Formants are tracked, and a 3D ball moves over a vowel-chart canvas in real time. Target spheres are placed at the target values of vowels, and the students’ task is to get the target spheres. A calibration process of capturing data from three cardinal vowels is used to normalize the effects of different size vocal tract, thus making it possible for people to use the program, regardless of age, size, or gender. A third formant is used in addition to the first and second formant, to distinguish the difference between two Swedish vowels.