Doing a actual walkthrough using info from a camera is many years away, and not even what I think would be the ultimate desired effect. In the short term, walking makes people vomit due to inner ear issues that nobody has resolved yet. Some have used "tunneling" which takes away your periphery and makes it more like a TV, to reduce the nausea. That could help here maybe. Teleporting from room to room and standing to look around is feasible with current tech, but those headsets are still $500-$800 and the cameras range from $200 - $60k. The average person won't have that lying around just to do walkthroughs. Not to mention you either need a $1000+ PC or a PS4 which brings the cost up even more. Google just announced Daydream, which is a headset you just drop your phone into. That's sub-$100. You still need to get their Pixel phone, which is $650. They have Google Cardboard, which works with just about any phone, but the quality is so bad, you wouldn't have a meaningful experience. In a few iterations (years), we'll likely have cheap cameras to generate the data and google glass like headsets on the cheap, or the ability to use just about any phone to view. That will be nice... but VR data generated from a camera still has other limitations, like the ability to look behind objects and outcroppings, since the cameras are stationary. You would definitely want to do that in a walkthrough.
I think tech like whatever the next iteration of Photosynth or 123D is, or other photos -> 3D apps, would solve that problem. You basically take photos of objects (or rooms) from many angles, and it automagically figures out how to construct those into a 3D space. You can fill in details by just taking more photos. Currently, it's a very manual process and doesn't do big spaces yet (that I know of), but if they get a camera that does something similar to the Google Maps shots then a realtor would walk through a house generating a bazillion pictures and munge it into a 3D house that a person could walk around in using a mouse/keyboard and controller. You could also save that to the web so anyone could do it in a browser on Redfin or whatever. You could also use that data in VR to walk around and experience at human scale, and allow you to look anywhere you wanted. Now we're talking! Because...
Since it's a 3D model, you could also enhance it. Someone could open it in 3D software and tag things with metadata from the MLS, so as you walk around you could interact to see "Roof, replaced in 2013" "Water Heater, brand new", "Living Room, 14x16" etc as you look around at objects. You could add an interactive overlay with school information or cap rates or whatever. You could stage it with stock 3D furniture or pictures on the wall, and allow the user to toggle "furnished" on and off. Maybe allow the user to change paint colors, or if it's easy enough to separate the appliances or cabinets, maybe they could swap those out to see what it would look like with stainless vs. white appliances. You could change the lighting. The options would be endless if it were a 3D model. All of this assumes streamlined software and iterated hardware to reduce the creation time, but this future is not far off.
You could do some of the tagging stuff with AR, right now if you are walking through the actual house. Just seeing overlays of metadata on the actual objects. You might have to tape up QR codes or something, but that could be a nice gimmick.
