Quantitative characterization of membrane defects (pores) is important for elucidating molecular basis of many membrane active peptides. We study kinetic defects induced by melittin in vesicular and planar lipid bilayers. Fluorescence spectroscopy measurements indicate that melittin induces time-dependent calcein leakage. Solution atomic force microscopy (AFM) is used to visualize melittin induced membrane defects. After initial equilibration, the most probable defect radius is ~3.8 nm in 1,2-dilauroyl-sn-glycero-3-phosphocholine (DLPC) bilayers. Unexpectedly, defects become larger with longer incubation, accompanied by substantial shape transformation. The initial defect radius is ~4.7 nm in 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) bilayers. Addition of 30mol% cholesterol to DOPC bilayers suppresses defect kinetics, although the inhibitory impact is negated by longer incubation. Overall, the kinetic rate of defect development follows DLPC > DOPC > DOPC/cholesterol. Kinetic defects are also observed when anionic lipids are present. Based on the observation that defects can occupy as large as 40% of the bilayer surface, we propose a kinetic defect growth model. We also study the effect of melittin on the phase behavior of DOPC/egg-sphingomyelin/cholesterol bilayers. We find that melittin initially suppresses or eliminates liquid-ordered (Lo) domains; Lo domains gradually emerge and become the dominant species with longer incubation. Defects in phase coexisting bilayers have a most probable radius of ~5 nm, and are exclusively localized in the liquid-disordered (Ld) phase. Our experimental data highlight that melittin induced membrane defects are not static; conversely, spontaneous defect growth is intrinsically associated with membrane permeabilization exerted by melittin.