Recordings have different durations

URL of experiment: Exp3_latest [PsychoPy]

Description of the problem: I ran this experiment online where each trial lasted 5.650 s:

  • a cross appeared on the screen for 0.5 s
  • a word appeared on the screen for 0.15 s
  • an image appeared and participants’ speech were recorded for 4.5 s.
  • There was 0.5 s of blank screen after the end of the recording before the next trial started

My issue, which I only realised after collecting all the data is that the duration of the recordings varies between 4.269909 s and 4.426644 s within participants (this is the case for all participants). I am interested in extracting exact reaction times so my questions are: has anyone experienced this before? And is there any way to know whether the recordings started late (cut at the beginning of the recording) or ended early (cut at the end)? If it ended early everytime, it is not an issue but if it started late everytime, my reaction times will not be accurate.

Thank you for your help!