We used simultaneous-multi-slice inverse imaging to characterize the relative latency and temporal variability of BOLD signals in the human ventral visual pathway with 0.1 s precision. The intra-subject and inter-subject variability were 0.39 (s) +/- 0.49 (s) and 0.51 (s) +/- 0.77 (s) when images of faces were presented. When scrambled faces were shown, the intra-subject and inter-subject variability were 0.42 (s) +/- 0.46 (s) and 0.56 (s) +/- 0.74 (s). With higher temporal resolutions, we showed temporal variability of HRF vary across cortical areas. Within the same cortical area, the temporal variability of HRF differ between different visual stimulations.