View Camera Mapping
Architecture Visualization in Motion Graphic Clips
CV 2006-2011: http://www.cineversity.com/forum/forum_posts.asp?TID=3660&PN=2
The first point which to clear is the amount of needed details. As you are working for a network and not for an architect, that might be differ a lot in terms of your amount work.
Is it a case for camera mapping? If so, do you have enough images?
If you need a high detailed model or a medium detailed model? Will say, what parts will be close to the camera and what parts will be more far away and can be replaced with textures, bump maps or displace maps.
If highly detailed, keep in mind that this building needs to be modeled only to a quarter, it repeats symmetrically on two axis and in that way it looks already less work. The mirroring can be done with standard C4D tools. But keep in mind that mirroring delivers huge data, especial for areas that the camera might not see. An object that is optimized for a specific camera view is mostly not mirrored fully.
If your level of detail (and the mix of “scale” so to say) is clear to you (Storyboard), the Camera movement should be considered, as details will be blurred during fast moves. A shallow DOF perhaps as well? (Yes, I have worked as a Set designer in feature films). The key is always to build for the resolution of the camera. Not for real-life.
So if textures will be a key element, you need a photographer.
If you send out a photographer, let me know so I can put list for together. Send me an e-mail if that is the case.
Most elements are repetitive and it is tempting to copy them (all), but I highly recommend to have a look what you really need!
Having several Architectural degrees and I have worked in an architecture offices as leading designer and office manger over years in Berlin. I’m happy to share my knowledge about Architectural problems during visualization. In your case it needs to be combined with the needs of Motion Graphic, and that combination is highly different to High End Architectural visualization, at least I assume it for now. (Mostly logos or words will be placed over the model over parts or complete over the clip time, which might take a lot of attention from the model.
A common mistake in that area is to model everything—just in case—and fill the scene with elements that will never be seen from the camera! Think more of a theater-stage than about the real world!
Hint: even if everything is modeled, you can re-project that high end model to a low end model. Camera mapping/projection is the key here. A few projections camera can cover a lot.
A rough storyboard would help tremendously.
Please allow a reduced list of points, reduced compared to the camera mapping series I try to release in 2012.
Set up a camera from you main view. Render a view, fairly high resolution, but note that you can do that always again and higher in resolution. Use a Protect Tag on you camera, just in case. Render always in 32bit/c, if anyone tells you 8 or 16 bit integer is fine, run!
Switch off all the main object layers, so it’s invisible for now.
What I like you to do is do model a cube like object with a roof, that looks like your object. Just really rough, but precise on the main edges.
Set up a material for that cube with roof and place the rendering in it, set the Texture Tag to Camera-mapping and drag the camera that was used for the rendering. Press the calculate button.
Now place a new camera close to the (projection) camera and explore how far you can move. In this way you can move everything in real-time, and the camera animation might be fun again.
Some areas, based on the animated camera might look distorted, make a copy of the animated camera from that point and delete the animation of it. Use this for a additional projection.
Check out some tutorials about the PMan option and apply that knowledge to the new camera, with an alpha-channel. In this way you can patch up that distorted area, The new “material tag” sits right sided from the previous one.
Again, Camera projection is just one more projection method. It takes the perspective distortion out of the image, based on a special geometry that fits to the main shape of the object
This is just a little practically drill to give you an idea of what is it all about.
CV 2006-2011 http://www.cineversity.com/forum/forum_posts.asp?TID=3605&PN=7
The simple geometry “stuff” is in the beginner tutorials to explain the tools and methods. Not to show that this is the standard idea behind it nor that this would be the border of options.
In the moment “you” take a photo, you merge four dimensions (space in time) into two dimensions, well time is frozen not squeezed, if you like.
Camera mapping is a technique to get the third dimension back and with a camera move to enable the fourth dimension time as well.
There are two methods, trial an error or re-engineer the camera point of view (POV) and camera point of interest (POI) as well the Field of View (FOV), the distance and banking of the camera needs to be established as well.
The idea behind showing it on simple objects is of course to give the attention to the technique, not to something that is just cool. We don’t work with such tricks here. Pure information and solid educational material is our target.
If the POI, POV, VOF and distance as well the banking is known (based on re-engineering or measuring on set, maybe both), then a simplified version of the objects can help to re-create partly the three dimensional space.
In the moment we take a photo, the objects become flat and with that a perspective “distortion” is introduced. If you project now from the POV to a geometry, which has the same constellation as the object/camera before, then the projection will help to get the distortion out of the image, based on the projection and the details given by the geometry.
In the case of a landscape, this is not doable with one shot, as the methods of perspective re-engineering will not work here. The trial and error method comes here closer. Again - if only one shot is available, it is difficult to get perfect results. If you have several shots with a similar POI but from different POVs, then several application will support an re-engineering process. Eight images is the rule of thumb, if the distance among the POV is substantial. Syntheyes is able to do such things…
On the other hand landscapes are relatively forgiving when it comes to trial and error methods. Which is normally given when you use a mesh and the magnet tool to re-create the topology of the landscape in the mesh, to un-distort the perspective distortion of it. Move the render camera constantly to measure your progress.
In a professional environment you might find, from time to time, that the camera has option to record the local position, which helps to put the camera into the landscape. With specific data called DEM (for example) you can get highly precise landscape data for most parts of the world. (C4D can read DEM files and there are many free ones available.
In your case (just one image), you need to evaluate the camera position and start with a sphere to explore the space when the rocks looks less distorted, from a different camera view than the Projection camera. Rule of thumb here, check if the image is cropped, if so, search for new stuff.
The POI with any normal lens is in the middle of the image and the camera was always (again normal lenses, e.g., no shift lens) perpendicular to the image. An eye for wide, normal, or long lenses should be given, to estimate roughly the distance to the image itself. This is most of the time easy to gain and the adanvatages on the long run weights the effort easily off.
If that relation is given, I would not do standard procedural for images first, I would first check out how far I can get with this stuff.
In my trailer (JET) you can find a scene where I have projected just little stones and other parts, to get the feeling of something bigger. It is like sculpting the landscape for the parts that you need.
I have done many landscape projections in this movie. It is printed in “Cinemascope 35mm Technicolor” and each scene worked on the big screen just fine. I tried to use as much different techniques as possible to show how it works in my making of movies.
If you have only one image, the process of “4d to 2d” occludes lots of parts in an image, which needs to be re-covered manually, will say: Photoshop’s Stamp-tool.
I use Camera Projection since over a decade. Eleven years ago I did my first documentary opener for a German TV station with it, and over 5 years ago, I did the first one week workshop about it. I like this technique so much that it is part of the current production of a music video (‘11), that I direct and produce (Making Of: here on CV of course, if interest exists)