Twelve student actors (6 males, 6 females) were given voice training for two months. Randomly selected, half of the students (3 males, 3 females) was trained in the traditional way, while the other half was given biofeedback with real-time spectrum analysis. The aim was a ringing voice quality with strong overtones at 3-5 kHz. Text samples read at different loudness levels were recorded before and after training. Fundamental frequency (F0), sound pressure level (SPL) and long-term-average spectrum (LTAS) analyses were made. Voice quality was evaluated by two voice trainers. Sound energy at 3-5 kHz increased by 3-4 dB (1.5-14.5 dB) across groups after training. This change, which was slightly larger for the biofeedback (BF) group, did not correlate with SPL. F0 increased slightly in the BF group and decreased in the control group. The relative dB level of fundamental decreased significantly more in the BF group probably suggesting a tighter adduction. Voice quality improved in both groups. Visual feedback seems to add some efficacy in voice training. However, there is a danger of hyperfunctional voice production if other sensory feedback is neglected.