A method, system, and medium for affective music recommendation and composition. A listener's current affective state and target affective state are identified, and an audio stream, such as a music playlist, is generated with the intent of effecting a controlled trajectory of the listener's affective state from the current state to the target state. The audio stream is generated by a machine learning system trained using data from the listener and/or other users indicating the effectiveness of specific audio segments, or audio segments having specific features, in effecting the desired affective trajectory. The audio stream is presented to the user as an auditory stimulus. The machine learning system may be updated based on the affective state changes induced in the listener after exposure to the auditory stimulus. The machine learning system may also be used to compose, master, and/or adapt music configured to induce specific affective responses in listeners.