Does anyone know how those crazy HP photography commercials were made (like the one with the guy singing and putting pictures down over his head and then bringing them back up)? They’re driving me nuts trying to figure it out!
My thought is that he starts out with a blue screen-type card, and when it gets close enough to his face, they stop, switch the card to one that is just an edge, start filming at the exact same spot, and visa versa.
I’m not sure what you’re talking about… but I don’t think bluescreens are used too much anymore with the power of CGI and all that stuff. I could be wrong… but I’d just guess it’s some fancy computer work (as is everything :P)
The commercial is about this guy, who holds these picture frames up to his face at different angles and whatnot, and then, when he removes them from his face, the image that the frame was surrounding is in the frame. I, however, have no idea how it is done.
I have not seen a write-up of the methods actually used by the HP commercial team (I’ve checked some of the trade press, but haven’t found anything published yet), so I can’t state categorically that I know the techniques that were used. But I have talked with several of my friends that are commercial animators, and we have figured out how we would do it if given the job. It actually is pretty straigt-forward:
Measure the distance from the camera to the actors, and the dimensions of the picture frames. Measure the intensity, color temperature, and position of the primary light sources. Save this info for later.
Film the commercial all the way through with full sets. The picture frames are just empty cardboard frames, painted white. No blue screen or stop action necessary.
Use CAD software to design a picture frame polygon, with the same exterior dimensions as the original picture frame, but no center hole (i.e. a solid polygon). Build a lighting model based on the information saved in Step 1.
In post editing, stop the each point where you want an “empty frame” to become a “frame with picture.” Capture the coordinates of each corner of the picture frame in the image, and save the image within this polygon to a separate location.
Using the information saved in Step 1 and the captured coordinates of the frame corners, solve to determine the position and orientation of the frame polygon relative to the camera. You can do this manually, or there are several commercial software packages that will do this automatically for every frame once you identify the initial polygon (note: this is a fairly common technique - you can find lots of information on extracting geometry from images in any of several image processing texts; one good place to start is with this Tutorial on 3-D Modeling From Images).
Adjust the relative postion and orientation of the CAD frame model and virutal camera to be identical to the real scene. Texture map the image saved in Step 4 on to the CAD polygon. Render the texture mapped picture frame with a 100% alpha channel for the image background. Composite this back on to the original image from the filmed piece.
Repeat Steps 4-6 for each version of the picture frame in the piece, always using the same image captured in the first run-through of Step 4.
That should do it!
-dave
Thank you for such a detailed answer Dave! That answered my question completely. :yikes:
Runs off to start an ‘I am Dave Lavery’ cult, er, group 
Just saw another one of these recently. They are just about the neatest commercial on TV, IMHO. 
You know the sad thing is I understand what Dave said and it was pretty much what I was going to say, there never is any pictures there, they take a screen from the video where the photo would be and use it as a skin for the model and have it move with the frame.