The difference is with testimony, the witness can be as moral and honest as possible but still miss report what they heard. With faking audio, it is either correct or someone purposely tried to fake evidence. That means that if the source of a recording could be verified, we know that it is very close to true. With witnesses, there is always doubt.
You're right to be skeptical of witnesses, but recordings are also not either/or. Even if you have some sort of digital signature scheme such that a camera can somehow sign its footage and the time and location it was recorded (a scheme that currently is fairly uncommon), how could it encode the circumstance in which the footage was recorded? Who is responsible for maintaining the signature scheme, and why can we trust them? In the absence of any aspect of that scheme, why shouldn't we closely scrutinize video and audio evidence as much as we scrutinize witness testimony?
The thing is that we have to believe someone when trying to understand what happened. A person, no matter how honest, can make mistakes. With a recording all that is necessary is to make sure that it is honest.
