Many empathy tasks lack ecological validity due to their use of simplistic stimuli and static analytical approaches. Empathic accuracy tasks overcome these limitations by using autobiographical emotional video clips. Usually, a single measure of empathic accuracy is computed by correlating the participants' continuous ratings of the narrator's emotional state with the narrator's own ratings. In this study, we validated a modified empathic accuracy task. A valence-independent rating of the narrator's emotional intensity was added to provide comparability between videos portraying different primary emotions and to explore changes in neural activity related to variations in emotional intensity over time. We also added a new neutral control condition to investigate general emotional processing. In the scanner, 34 healthy participants watched 6 video clips of people talking about an autobiographical event (2 sad, 2 happy and 2 neutral clips) while continuously rating the narrator's emotional intensity. Fluctuation in perceived emotional intensity correlated with activity in brain regions previously implicated in cognitive empathy (bilateral superior temporal sulcus, temporoparietal junction, and temporal pole) and affective empathy (right anterior insula and inferior frontal gyrus). When emotional video clips were compared to neutral video clips, we observed higher activity in similar brain regions. Empathic accuracy, on the other hand, was only positively related to activation in regions that have been implicated in cognitive empathy. Our modified empathic accuracy task provides a new method for studying the underlying components and dynamic processes involved in empathy. While the task elicited both cognitive and affective empathy, successful tracking of others' emotions relied predominantly on the cognitive components of empathy. The fMRI data analysis techniques developed here may prove valuable in characterising the neural basis of empathic difficulties observed across a range of psychiatric conditions.