Systems and techniques are provided for occupancy based volume adjustment. A signal including detected locations for several persons may be received. An occupancy model may be generated based on the detected locations. A volume adjustment for a speaker may be generated based on the occupancy model and a target sound level range for each of the detected locations. The volume of the speaker may be adjusted based on the volume adjustment. A signal including a detected location of one other person may be received. A location of other speakers may be received. The occupancy model may be generated based on the detected location of the one other person. The speaker may be determined to be closer to the one other person than any of the other speakers. The volume adjustment for the speaker may be generated based in part on the location of the one other person.