The ability of preventing unwanted memory from invading awareness is important for mental health. Using the Think/No-Think paradigm, previous studies have proved that individuals can forget episodic memory by direct suppressing them repeatedly. Our unwanted memories come from multiple sensory channels in daily life. However, most researches adopted visual materials, and few researches considered other sensory domains. In the present study, we used auditory-visual materials to investigate the cross-modal memory inhibition, and separated the control instruction and memory cue to test the anticipatory effect. Thirty-one healthy participants who were native speakers of Chinese were recruited from Southwest University, 19 of whom were female. And their average age is 20 (SD=1.73). The stimuli consisted of 64 randomly combined sound-word pairs, and 16 of them were used for practice. The sound materials with 2.5s long were taken from common sounds in daily life, such as dog barking and motorcycle launching. The word materials were nouns with two Chinese characters. The procedure included learning phase, pre-test, Think/No-Think phase and post-test phase. In the learning phase, the participants were required to first passively memorize and then actively learn with feedback. In the Think/No-Think phase, the memory cue was played for 2.5s, followed by a 2.5s blank. The Think/No-Think control instruction was presented with 1s before the memory cue (Hanslmayr et al., 2009). There were 12 repetitions for each trial, and EEG was recorded during the Think/No-think phase. Both the same probe and the independent probe were tested after the Think/No-think phase. We found that the recall performance of suppression condition is significantly worse than that of baseline condition in the independent probe rather than the same probe. Event-related potentials (ERPs) revealed two main findings: 1) in the last 6 repetitions, negativity in the right frontal-central area is enhanced by No-think instruction compared with Think instruction before memory cue presented (300~1000ms); and the difference between two conditions is due to the negative enhancement under the No-think instruction instead of the Think instruction from the first repetition set to the second repetition set; 2) during 500~1500 ms after memory cue presented, positive potential in the central-parietal area related to No-think items is less than that related to Think items, while activity in the right frontal area is reversed; and compared to unlearned items, learned items with No-think instruction induce more positive activity in the right frontal area, while there is no significant difference between learned and unlearned items under Think instruction. The results demonstrated that the active forgetting effect can be achieved on target memory itself with the auditory-visual cross-modal condition. And we verified there was an expected processing before memory suppression if the task instruction was presented in advance. In addition, we found that the continuous inhibition control process under the condition of memory suppression triggers the late positive potentials in the right front. Further researches will need to explore the sources of anticipation and the late inhibition control components, and adopt a variety of sensory materials on the study of cross-modal memory suppression to promote the clinical applicability of the active forgetting effect.