Some of us are old enough to remember when video was an exciting new tool for researchers in the social sciences, particularly in the area of social interaction, and especially where the interactions were largely non-verbal. Because it allows a repeated review of the same visual sequence, you are enabled to see the fine-grained elements of these interactions, and their complexity.
I can recollect sitting with a PhD student who was researching mother– baby interaction, focusing on the analysis of sequences of mutual imitation – one of the earliest forms of communication. The student played through an edited section of the tape several times; at first I couldn't 'see' what was going on. But gradually, and then quite clearly, I did.
There is, of course, a big difference in the communicative capacity of a sixmonth-old baby with its mother, and two adults in an interview situation. But non-verbal aspects of communication between people who are talking to each other are an important, because qualifying, aspect of the meaning of what is being said.
By definition, an audio recording loses those aspects of communication which are non-verbal, which means that a layer of meaning is stripped out. But the process of transcription removes another layer: those paralinguistic features such as tone, pace and emphasis that further qualify the actual words people are using. Reading the written-word version of an interview, even when faithfully transcribed, you can sometimes feel (especially if you conducted the live interview) that it has somehow been de-natured.
What follows from this is that a video recording of an interview is the most complete account, however it is subsequently dealt with.