A few notes on creating a giantess
July 5, 1998
The art of making collages used to require very careful hand work with a razor blade or X-Acto knife, an air brush, and considerable artistic talent. People have been creating giantess collages for years with some of the most successful work appearing in BF Goodrich ads in the mid 1960s as well as on the cover of Stern magazine during roughly the same period.
Digital darkroom tools are now quite powerful and the required hardware platforms have fallen dramatically in price. Until 8-95 my platform consisted of PS 3.0.1 running on an Apple Centris 660av with 36 MB of RAM. I am now using a Power Macintosh 9500/250 G3 (I broke down and bought a 250 MHz Newer Technology G3 processor card) with 144 MB of RAM and Photoshop 5.0. I *really* like this platform - it is fast and extremely stable. I usually allocate about 100 MB to Photoshop and am not running out of room even on the most complex work I've attempted (of course this can always change:-) The next change will be a processor upgrade as some of the filters require several minutes to execute.
You will need a good monitor. I have a fifteen inch Sony and drive it with 24 bits/pixel. I dislike larger monitors good ones tend to cost an enormous amount and less expensive ones suffer from edge distortion. Since I tend to work in very small regions this isn't a big deal.
Be sure whatever graphics board you are using (or internal color) is good. This seems to be a very weak area on many Windows machines and you need to minimize color mapping problems.
I have a nice graphics tablet, but tend to use it mainly for freehand drawing (my average collage has quite a bit of freehand work). For the bit level work I still use a mouse most of the time. I don't like Apple mice and have gone with the Kensington mouse, which is much smoother and has less hysteresis. Also recommended is a good mouse pad - I like the 3M Precise Mousing Surface.
Don't let heavy hardware requirements scare you away. You can get started with very modest hardware. My first two hundred collages were done on the Centris starting with just 20MB of RAM and working to 36MB. My all time favorite piece of work was done on this machine. As your technique improves, you'll need more memory as well as horsepower, but you can start learning with much less. Also note that memory is very cheap these days!
Around September 1995 it became clear that some custom filters and plugins were going to be necessary and I began to play with homemade Photoshop plugins. It isn't an impossible thing to do, but is reasonably geekish - there are many image processing books/journals in the EE library of your local University. If you are comfortable with programming there are any number of tricks that transcend what is available in "stock" PS. There are also a number of third party plugins, although most of these are for wild effects that aren't all that useful to me (you do need the Kai plugins even if you don't use them because they are so much fun!).
I don't want to get into operating system wars, but I have been very pleased with the Power Macintosh. It is reasonably fast and easy to deal with -- OS upgrades are trivial, devices like printers and scanners are easy to deal with and it doesn't crash. People claim the Mac is not a programmer's machine, but the Codewarrior C++ compiler is great for building non-trivial plugins. If you go this route be sure to grab the Adobe plugin SDK from their website.
Enough of the hardware and software! You really need to spend some time to develop your "eye." If you don't sketch I recommend getting a few good "how to draw" books and practice practice practice. Stick with it and you will get reasonably good in a few months. Any attempts to play with computer art before you develop a basic artistic sense will be wasted. I find working in pencil and charcoal to be extremely revealing. Sometimes I print out backgrounds and sketch models directly into them to work out ideas.
Good input images are essential to a believable collage. I find the following qualities important: paper quality, print quality, lighting angles, model point of view (PoV), color saturation, model pose, scene size (is the model cropped?) and general model type. Now you must find another photo or set of photos to add to the collage. They must also have similar qualities. Finding reasonable matches has been (in my experience) impossible, so one is forced to compromise or hire real models who can be posed by the photographer (I'm not that professional -- if anyone knows of a model who wants to be really big I can probably help her :-)
Once you've located the appropriate images it is necessary to convert them to a digital form. I have just replaced my old HP IIcx scanner with a Linotype-Hell Jade and am thrilled with the initial results. The old HP performed well for several years, but was clearly outdated and the Jade is far and away the best unit I've seen in its price range. I also use Kodak PhotoCDs, which have impressive quality if your photography is good. Kodak also offers a professional version which seems more applicable to medium format than 35mm photography.
It has been my experience that few magazines print well enough to justify anything beyond 200 by 200 dpi scan settings. It turns out that a 200 by 200 dpi scan of a normal magazine page will result in a Photoshop image file that exceeds 10 megabytes, so denser scans are only for people with huge amounts of memory as well as high quality input images.
Scan the images at the highest density your equipment and the subject permits. For me this is either 150 by 150 dpi or 200 by 200 dpi. Even though I optimize my work for computer screens, high density scans are very important as you will see. My scanner allows 24 bits per pixel and I always use this even for black and white as it gives me greater latitude than the 8 bit black and white mode. Different scanners may behave differently.
Now that you have your two more pictures it is time to set color balances, contrast, darkness, etc to match the scenes as closely as possible. Fine tuning takes place one the transfer has been made, but this is a good time to get the gross effects. Often you will notice that a collage isn't possible at this stage and you will avoid wasting considerable time if this is discovered early in the process. I find it useful to make a few freehand sketches of what is wanted to see if shadowing will work and get a sense of luminosity requirements.
Now for the fun part :-) Take the woman's picture and take out everything that doesn't look like a giantess. Photoshop offers some tools for accomplishing this, but aliasing by the program looks funny and you have to do much of the background subtraction on a pixel by pixel basis.. You may have to deal with print on top of the model (from an ad for example) or missing features like the top of the head, etc. Here is where you have to sketch in bits and pieces. Depending on the subject the process can be laborious.
I find that supersampling is often necessary - if you are at 200 dpi take a troublesome portion of an image and turn it into a 400 x 400 or denser image. Now the editing is much easier and subtle effects can be generated. When you are finished you move it back to 200 x 200 (or whatever) and portions of the effects are still present. Photoshop aliasing is robust, but this is one area where often use some homemade tools.
When the giantess to be is ready for transfer you can move to the background photo and create regions that will be in the foreground. Copy the giantess and paste her behind these regions. Now fine tune the position, her size, and lighting effects before deselecting her. It find it useful to toggle between the Photoshop "show boundaries" settings here.
This is the appropriate place to play around with perspective (you are working in layers aren't you?) of both scenes and do a bit of image warping to create the illusion of a match. This is probably the most difficult (at least for me) part of the process and is the area where I was forced to create a few tools.
When you are finally satisfied, or (as in my case) are sufficiently frustrated that you're ready to give up and move on to something else, you can finally merge the layers into a single unit.
You can now work with her boundaries again on a pixel by pixel basis. Using partial transparencies on brushes, the smudge tools and the copy tool turns out to be effective and freehand sketching is usually necessary.
You can now draw in shadows by hand (rarely can you use a drop shadow - in fact I have never experienced good results from any of the shadowing tools) and tune the overall distortion of the picture - you can experiment will image sharpening and blurring locally (I find gaussian distortion and Lab color channels particularly effective). Adding a bit of noise sometimes works (particularly on black and white collages), but all of this depends on the subject at hand.
Spend some time with the burn and dodge tools as well as some more freehand drawing to tweak up the image.
Sit back and admire what you have -- it is now time to reduce it to match screen resolutions as well as shrink the file size. Of course if you were going for print, the next steps would be inappropriate as you are now at maximum image quality. I usually use 100 dpi settings. Photoshop is very good at the interpolation and the net effect is that some of the edge problems you've been worried about become invisible. Working at 1.5 or more times the target pixel density seems to work well for most of my collages.
Photoshop is a lossless format, but it is very large. I save most of my work at 100 dpi and then experiment with jpeg settings. Sometimes the jpeg compression actually makes the collage look better on a computer screen, so try many settings! It turns out that different programs save jpegs differently. Sometimes I use DeBabalizer if Photoshop isn't doing exactly what I want.
The issue of targeting to specific platforms raises its ugly head as graphics boards in PCs differ wildly and it is unlikely that your collage will look the same across platforms. Macs are very good at producing predictable results that look the same on other Macs (modulo screen depth), which is one reason I use that platform. I can't stress enough that you need more than 8 bits of color depth and you should resist the urge to reduce the color map from 16 or 24 bits per pixel. This creates a real problem if you want to display on the web as the average web user looks at images from the internal viewers in Netscape or Internet Explorer, both of which give terrible results. I choose to ignore this issue and recommend that people use quality 24 bit displays and viewers as "helper applications."
Some tricks - European fashion magazines (except for English magazines) are printed on good paper with superior resolution. The Europeans also have a tendency not to crop the subjects (which is very common in the US). If you're lucky you'll find something photographed against a mono-colored background, which makes the image extraction much easier.
When working with your own images I have found glossy prints to work best. I usually have them printed at 5 x 7 or 8 x 10 inch sizes specifying a glossy print. These generally can be scanned at 200+ dpi and I have recently been scanning at 400 dpi, although you clearly need memory to do this.
I have also been working with handdrawn backgrounds, stylized backgrounds and sketched images recently. Although not photo realistic, this style allows a much greater degree of freedom. It is interesting, but I personally don't enjoy the process as much as my other work.
I have not gone into important features like multiple layers and only note there are numerous other Photoshop tricks that are probably best learned by experimentation. Most of my collages tend to be experiments playing with some PS feature that I haven't mastered or a new plugin I'm trying to get working.