We already have neural networks that can create images that roughly resemble what a person is looking at.

I don't think it would entirely be out of the question for people to be able make videos from thought when asked to recall a memory.

However, although the subject and location of these videos would often be correct, I doubt the smaller details in these videos would be at all consistantly accurate or exist at all. I highly doubt that you would be able to scan a person's memory for clues like in some sci-fi movie, because ultimately memory is flawed and peoples brains will fabricate details to fill in the gaps.

Short answer, yes but it won't be a perfect copy of what happened.