Understanding speech masked by a competing talker, namely, speech-on-speech perception, remains a difficult listening situation for users of cochlear implants (CIs). The performance shows large variability amongst CI users, implying that the CI device may provide sufficient speech cues for the task and as a result, training may help increase the proficiency level. Previous studies with musician populations indicated that music training, via transfer of learning, can provide an advantage for speech-on-speech. In the present study, we used music-based training, a dedicated piano method developed to be suitable for CI users— Guided Audiomotor Exploration (GAME). GAME training is based on finger movements on the keyboard and extemporized combinations of short musical structures. Emphasis on finger movements helps stimulate motor neural networks as part of audiomotor integration, reinforced by improvisation exercises. Further, social interactions with an instructor or in a group setting seem to provide an enjoyable training experience. Twenty-four CI participants were (pseudo-)randomly assigned to one of the three groups: the GAME training for 6 months, a control group with Minecraft lessons, and a control group of no training (only test effects). Minecraft lessons for the first control group were designed and delivered in a way similar to the piano lessons, involving social interaction with a teacher as well as the motor component, but lacking the complex audio feedback involved in music making. Participants were tested before and after training, in a speech-on-speech task based on the coordinate response measure. Listeners had to identify a color and number in a sentence presented simultaneously with another speech stream from the same talker, with the same or altered voice parameters. The results show that the GAME training had a significant positive effect on speech-on-speech perception compared to no-training. In contrast, the Minecraft training showed some small improvement, but this was not significantly different from the no-training group. These results indicate that this type of musical training could be beneficial for CI recipients.