The Concept

Before we get into implementation details, let's just go over how things went from the client's perspective.  We shot portraits before and after our Saturday and Sunday services, for a total of about 90 families.  We had a table set up apart from the actual portrait setup where volunteers would hand incoming families a copyright release letter (so they can print their photos without issue) and a business card sized slip with a URL, a unique alphanumeric code, and a matching QR code printed on it.  That card was what made the magic happen, and we'll go into them in detail soon.

An example of a client card.

From there, the greeters directed them to one of the photographers: there were two of us working, one indoors and one outdoors.  Clients were free to visit both, but most just went to one or the other.  When one of them actually got to a photographer, we would snap a photo of their card, making sure the QR code was clearly visible, and then just go ahead and shoot the photos of them.  Starting the next day, they were able to visit the URL printed on the card, type in their code, and view/download their photos.  So, how did we get this all to work so smoothly?

Step One: Generating the Cards

Of course, the very first thing to do was to get those cards printed.  To actually generate the cards, I wrote a script called gencards.py.  If you're following along at home, you'll find it in the scripts/ directory of my repository.  This script is run from the command line and accepts two arguments: the number of cards to generate and the file to write them to.  Give it a few seconds to run, and it spits out a PDF that you can print with your desired number of cards, each with a randomly generated code and a matching QR code.  Just about every aspect of the generator is customizable, but you'll have to edit the script to change them.  If you open it up in a text editor, at the top you'll find a section labeled "User Parameters."  By changing the values in that section, you can change the URL printed at the top of the card, the positioning of the card elements, the page size, and some other details.  gencards.py uses a program called qrencode to generate the QR codes, so you'll need this program installed and in your system path for the script to work.

Luckily enough, we have a staff member who works for a print shop, so he was able to get a couple hundred of these cards printed up easily enough.  With a box full of cards (and the copyright release letters), we were all ready to go with our printed material.  From there I just needed the software to handle post-processing after the event.

Step Two: Splitting up the Photos

When we actually shot the photos, the crucial step was snapping a shot of each client's card before starting their session.  Afterwards, a script I wrote called sortphotos.py was able to comb through them and split them up into directories by their codes.

An example of a card shot.

sortphoto.py takes just two command-line arguments, the input directory and the output directory.  It goes through every photo in the input directory (and any subdirectory of it), and copies the photos it finds into the output directory split up by the code cards that come before them.  It uses a program called ZBar to actually read the QR codes, and ImageMagick to resize images prior to reading, so you'll need to ensure that both of those programs are installed and in your system path for it to work.

It's also capable of handling RAW files, using ufraw-batch to do the RAW conversions.  To make this work, obviously, you'll need ufraw-batch installed.  To control the RAW conversion, you'll want to load one of your RAW files in Ufraw and save a .ufraw file with the settings you want to use.  Just put the .ufraw file in your input directory, and it will automatically be used for all the RAW files in the directory.  If you want to use different settings for different files, you can put them in different subdirectories with ufraw files in each of them, and the script will always use the one closest in the directory tree to the RAW file.

Step Three: Generating Thumbnails

In our case, this was actually step four, but hopefully you'll learn from my mistakes and do it third.  If you're going to post your freshly sorted photos on the web, you're probably going to want smaller thumbnail sized images to display.  The viewer script that I wrote to display the images online will generate thumbnails if none are present, but it's oblivious to the EXIF rotation attribute so vertical images may appear sideways, and if there are too many or too large of images (this was our problem) it could run up against PHP's memory or time limits and crash.  To solve that problem (very frantically after uploading all the photos and finding them not displaying, I might add), I wrote a simple script that uses ImageMagick to generate thumbnails ahead of time.  It improves site responsiveness, it avoids potential PHP crashes, and it makes better looking thumbnails to boot.

The script is called genthumbs.py, and it accepts three command-line arguments: the input directory, the maximum width, and the maximum height.  The script then finds all the images in the input directory and its subdirectories, and for each one generates a thumbnail image with width and height no larger than the given maximums.  It names the thumbnail originalfilename_thumb.jpg, so for instance a thumbnail of img1.jpg would be img1_thumb.jpg.  This is the convention that the web viewer expects, so the generated thumbnails will work flawlessly with it.

Step Four: Uploading and Viewing

After the images were sorted and the thumbnails generated, the last step was to host them on the web so clients could retrieve their images.  For that, I wrote a very simple PHP script that you'll find in scripts/web/ in my repository as index.php.  The design is no work of art, and the source code is far from perfect, but it's a quick script that works well enough.  If you're using the system professionally, you'll probably want to improve the web viewer script (and if you do, feel free to make a merge request with my repository, I'd love to add your improvements).

Assuming that the viewer as-is will work well enough for your needs, you just need to put both it and the sorted photos up on your web server where they're visible by the public.  You'll need to change two constants at the top of the viewer script, PHOTOS_LOCAL and PHOTOS_WEB, to make it work correctly.  Set PHOTOS_LOCAL to the path on your server to the photos directory and PHOTOS_WEB to the URL that points at the root of your photos directory.  In most cases these will probably be the same (when we deployed it they were), but having them separated gives you a little bit of extra flexibility if, for instance, the two are being served from different domains on the same server.

Once the viewer is up, typing a code with images associated with it into the text box will bring up a page showing thumbnails of the images with links to the full size files.  You can also add &index=t to the end of the page URL to include the card shots along with the actual photos (they're normally not shown to the user).  If the user enters an invalid code, they'll see a friendly error message and a prompt to try again.

Conclusion

Aside from a small panic when I realized that my viewer script was crashing while generating thumbnails, everything went wonderfully smooth.  We got everyone photographed and the photos posted and sorted without too much effort, and we used no special hardware or commercial software in the process.  I realize that operating these scripts may be a little complicated for some, but I'm hoping that if you're in a similar situation to us they can help you out.  If you're part of an organization trying to offer a similar service and having trouble getting anything to work, feel free to contact me with questions.  And of course if you make any improvements to my scripts, I'd love to see them!

Update - Per a suggestion I received on Github, I've changed the standard filename for web viewer thumbnails from thumbfilename.jpg to filename_thumb.jpg.  If you already have thumbnails generated and you change to the new version of the viewer script, you'll need to make sure to regenerate your thumbnails.

The mod is relatively simple, but you'll need basic soldering skills.  In addition, my method removes the auto flash sensor to fit in the 3.5mm jack, so you won't be able to use the auto mode afterwards.  Of course, messing around inside a speedlight can be dangerous, so I'll start out with the obligatory disclaimer.

WARNING: The inside of your flash contains a very large, very dangerous capacitor.  Take caution opening and modifying the flash.  You are solely responsible for any damage you cause to your flash or yourself.

If you want to do this yourself, you'll want to start out by discharging your flash as thoroughly as possible.  The capacitor inside is exceptionally dangerous, so you don't want it charged while you're working.  I just put the flash in manual mode and set it to full power.  I slide the battery door down but hold it in place, pop the flash and release the battery door immediately afterwards to keep it from charging back up.  Once you've got the batteries out, turn the flash head sideways to get access to the screws that hold the body together.  You need to undo two screws on the same side of the body (on opposite sides of the head), it doesn't matter which two.

Once you've got those two screws undone, you can turn the flash over and unscrew the four screws attaching the body to the foot.

With those undone, the two halves of the body will be free to seperate.  Pull the front of the body off.  The battery door will come off, just set it aside for now.

You'll see several cables attaching the front of the shell to the insides of the flash.  Disconnect two of them: the one that attaches to the middle of the front of the circuit board and the one that attaches to the bottom of the circuit board.  The other cables will be very tricky to replace if disconnected, so leave them in place.

With the front shell partially disconnected, you can now remove the foot entirely.  There are no screws involved, it will just click out of place.

Now pop the red piece of plastic off of the front shell.  From inside you can easily push its edges out with a screwdriver, they'll pop out if you push them inwards just a little bit.

That glass dome you see in the front is the sensor for the auto flash mode.  To remove it, just undo the three screws on the front and it will come right out.  This is where we'll install our 3.5mm jack later.

Next you'll need to drill a hole in the red plastic piece for the 3.5mm jack to fit through.

Drill until the hole is just big enough to fit the jack through, and then set the plastic piece aside.  For the next step, you'll want to reference the pinout for the Nikon hot shoe (second diagram in the linked forum post).  Looking down on the circuit board from the inside, the pins of interest will be the center pin (trigger), the lower right pin (quench), and the tab to the upper right (ground).  Desolder each of these pins, making absolutely sure that you have the right ones first.

Once you have them desoldered, you need to cut three wires long enough to make it from the foot up and out through the body (I cut mine about five or six inches long to start with).  Each wire will need to be soldered to one pin.

The pins on the circuit board itself have room enough around them to slide one end of a wire into the hole with them and solder the whole thing back together.  For the tab I just made a small loop of wire and wedged it in between the tab and the circuit board before soldering it back together.  When it's all soldered together, you should have all three wires firmly attached to their respective pins and the circuit board.  I used red for trigger, green for quench, and black for ground.

Once the wires are attached, plug the shoe back in to the bottom of the flash's main circuit board and press the shoe into place.

Reassemble the entire shell, running the wires out through the hole in the front.  You may want to route the wires to one side or the other going up from the foot into the body so that they don't get crushed by the small circuit board mounted on the front shell.  Don't forget to put the battery door back in its place before pressing the two sides of the shell together.  Also note that if the flash head at any point became dislodged, you'll need to hold down the horizontal locking button and press it firmly back into place before you'll be able to get the shell back together.

At this point I cut the wires so that I had about an inch and a half to work with out the front of the flash, then stripped them, tinned them, and bent the ends into hooks.  You'll need to connect these wires to your 3.5mm jack: ground goes to the base of the plug, trigger goes to the tip, and quench goes to the middle.  I strongly recommend just hooking the wires loosely into the jack and testing the entire unit before soldering anything.

Remember that you must put the flash in TTL mode in order for remote power control to work.  If the wires are connected properly then you should be able to control the flash's power with the dial for the receiver's group on your transmitter.

If everything is working correctly, then solder the wires to the 3.5mm jack.  Then just push the front of the jack through the hole in the plastic plate, screw its washer in on top of it, and then pop the plastic plate into place on the front of the flash.

If everything is assembled correctly, you can now plug your flash directly into a Radiopopper JrX receiver with any 3.5mm cable and get remote power control and triggering.  Be careful yanking the cable out of the flash, as the plastic plate may come out with it if you're not gentle enough.  If you're really worried about this, you can superglue it to the flash body, but I prefer to be able to open it in case I ever need to get in and repair something in a hurry (this actually happened with the first flash I did this to).

This technique will work with the SB-25 and SB-26, and I wouldn't be surprised to find out that the SB-24 has a similar internal structure, although I've never actually seen one before.  The same general technique should work for just about any Nikon flash up to the SB-800, but figuring out exactly where to put in the 3.5mm jack is left as an exercise to the reader.  Note that this technique will not work with Canon flashes as-is, because they use a different type of TTL control signal.  There is a circuit floating around that will allow you to connect a Canon flash directly to a Radiopopper JrX, but I don't know if it would be feasible to construct the whole thing inside the flash itself: you might be stuck with an adapter.

Starting Out

Lewis Park is located near the house where I grew up, and I met my friend Kelsey there to shoot a couple of days ago.  We met up right about an hour before sunset, which gave us some time to work in the golden light that you get while the sun is going down.  I started out by parking her right in front of a sun-beam streaming through the trees, putting an umbrella up on her opposite side, and doing some simple portraits with just the sun and the one light.  I put a 1/2 CTO gel on the keylight to better match her face to the golden sunlight, and I kept that gel on that light throughout the shoot.

I always try to start out a shoot with some standard portraits against the nicest backdrop I can find at the location.  They're simple enough that I know I'm guaranteed at least a couple of good solid images right from the beginning, and it's a good exercise to get me into the swing of manipulating my lights, get a feel for the ambient light, and so on.  In this case, I found that the sun was going down a little faster than I expected, and everything under the trees was in really heavy shadow.  In the photo above, you can see that the background is barely visible, while the bright sun ray is almost blowing out her hair.  For a couple of days I've been planning out a shot that involves some motion blur, and I was going to save it until the end when the sun went all the way down, but seeing how dark the shade was I decided to go for it right away.

Round and Round

I recently saw a long-exposure of a spinning merry-go-round, and I thought it would be really fun to shoot something like that...except from a perspective on the merry-go-round.  I started out testing the concept by trying to put a tripod on the merry-go-round.  It kind of worked, but the shot was insanely tight: this is one of those times when I really wish I had a lens shorter than 28mm.  In the end, I decided that I'd just have to hand-hold the shot from on the merry-go-round and do as best I could to get a reasonable field of view.  Once I'd verified that I could, in fact, get some decent background blur behind her, I set up some lights.  I put an SB-25 in a snoot to the left, and an SB-26 in an umbrella to the right for some fill.

Once I had the lights set up, I got the merry-go-round spinning, jumped on it, and then crouched down and wrapped my arm around one of the horses' necks, hanging on the outside of the horse.  Then I worked my camera up to my eye with my free hand, and focused and triggered the shutter with the same hand that was holding me on to the horse.  It was all at once nauseating and frightening.  I opened the shutter while Kelsey was around the area where the tree is in that setup shot, and it closed probably 1/4 turn after that.  The strobes gave me a nice crisp shot of Kelsey while the background blurred into streaks of motion behind her.  While a gorrila-pod and a wider lens would have made this shot a whole lot easier, I managed to make it work with just a tiny bit of ingenuity and a less-than-fully-formed self-preservation instinct.

Wrapping Up

After the merry-go-round shot, we decided to go for one final set of portraits on a nearby tree, which culminated in the photo at the top of the page.  The tree slopes up and over a little stream at a reasonable angle, so Kelsey was able to get up it without too much trouble.  At this point we were well into twilight, so I pushed my ISO up to 800 (I normally keep it at 200) and started looking for a shutter speed that would give me a reasonable amount of background detail.  I ended up settling on 1/30s, which was really pushing it even at the wide end of my 70-200, but my strobes froze my subject easily enough, and I managed to hold steady enough to keep the background detail from blurring horribly.

Of course, once she was up in the tree, I had to figure out just how I was going to light her.  My light stands were too short; I could have lit her from below, but then she'd have that creepy inverted shadow look on her face, and I didn't want to do that.  Getting a light stand up in the tree was out of the question, but I realized that I could probably get a single strobe to stick up there with a ball-bungee.  I put a couple of them in my kit years ago after I read Strobist's Lighting 101, and since then I've never found a situation where it actually helped to ball-bungee a flash to anything.  On this shoot, they finally got their day in the sun.

I climbed on up the tree and sure enough, there was a nice skinny branch sticking out of a limb right in front of my subject...it was as if Mother Nature decided to put a light stand there for me.  The branch was exactly the right size to hold a flash securely with a ball-bungee, so I attached the light, made sure to set it to its widest zoom setting so I didn't have to worry about aiming it precisely, and climbed back on down the tree.  I used a second light to backlight Kelsey, but I put that one on the ground behind the branch on a fully extended light stand and pointed it up at her.  The angle of the backlight isn't nearly so important as the key light.

If you look closely in the setup shot there, you can see the ball-bungeed strobe on the branch to the right, and the stand directly underneath Kelsey.  You may need to view it larger to see where they both are.  With the lights in place, I just adjusted their levels until I had a look that I liked, and started shooting.  This is the kind of situation where I'm earnestly thankful for the ability to remotely control my power levels with the Radiopopper JrX triggers.  Without them, changing the light levels would have meant crossing the stream to get to the backlight and climbing the tree to change the keylight.  Remote power control means that I can set my lights up in crazy positions like this and still dial in the power just where I want it.

After finishing those shots, we decided that between the sun going down and the temperature dropping (we're not accustomed to cold here in Florida), it was time to pack up.  So I got my light down from the tree, packed up all the stands, and we parted ways.  I ended up with some photographs that I'm really happy with, although unfortunately I lost my last umbrella in the line of duty.  A light stand blew over with it, and the fall finally killed it.  Fitting, I suppose.  While I tempted fate clinging to the outside of spinning metal disks and scaling trees, my trusty umbrella paid the ultimate price.

For this year's Christmas portrait, I put my girlfriend inside a snow globe.  In the last entry I talked about the hard part: photographing the snow globe.  Today's entry covers photographing Danie and actually getting her into the globe, which was really more drudgery than anything.

The Shot

The photo of Danie was pretty straightforward.  I just posed her on some seamless paper so that she would fit into the scene with the figurines, threw up a shoot-through umbrella on the left, and pointed another light up at the ceiling for fill.  Aside from the cat coming and messing up the background paper, it was just a matter of shooting until I got a pose that I thought would fit in well.

If I were shooting a standalone portrait, I'd put a lot more attention into the lighting, but I had basically no room to work with in my cramped living room, and I knew it was going to get shrunken down to a tiny size anyways.  I'd normally be trying to get the background just right, but since it was just going to be cut out in the end, why bother?

The Compositing

Once I had the shot I wanted to use, I had to composite the two together.  I made a layer mask to get Danie off of her background, which was just mind-numbingly tedious, as layer masking always is.  Once I had her masked out, I went over to my snow globe image and made a new layer and copy/pasted the figurines in the area she would go onto it.  I made a mask for it, showing only the figurines that I wanted to be in front of her, and then dropped her into that image, and put her layer right in between the foreground figurines and the background.  Then I just scaled her to the appropriate size, and moved her into the position I wanted.  With all that done, my 2010 Christmas portrait is complete.  Next year, I'm thinking I may go for something a little less time-consuming.

The Setup

I've detailed a dark-frame glass setup before, and what I did this time was the opposite: a bright-field glass shot.  The background was white with a black border just out of frame to refract in the edges of the glass, giving it those dark edges to define its form.  To get that effect, I took a sheet of black posterboard and just cut a rectangular hole right in the middle of it, slightly larger than the snowglobe.  I super-glued some white copy paper in its place, and then lit it from behind with a snooted speedlight for a nice white background with black edges.  I hung that background from the crossbar of my backdrop stand with some gaffer tape, and piled together some tupperware bins to get the snow globe up to its level.  For my exposure, I just set my camera to ISO 100, 1/250 (my max flash sync speed) and f/5.6 (to get a little bit of depth of field inside the globe) and then brought the background light's power level up until the background was just a tiny bit short of blowing out, pushed right up on the right side of the histogram.

The image above shows the backdrop-only photograph (ignore the mottling in the background, that only came out because I exported this one directly from my photo manager, and it applied its own exposure compensation), which is where I would normally leave a bright-field glass shot.  It's got some nice thick lines there to define the glass' form, but in this case there's also a lot of foreground detail that's been lost to the silhouetting, so I needed to add a front light.  Of course I don't want a bare speedlight, I need a light source a little bigger and softer.  This is where I would normally reach for my trusty umbrella, but for glass that's a no-go.

An umbrella makes a great portable soft light source, but when you shoot glass it reflects whatever light source you point at it right back at you.  In the case of an umbrella, that means the odd shape and the spokes all show up in your image, which is not pretty.  For glass you want something rectangular, so you get nice rectangular highlights, highlights that could pass for window-light.  Yesterday I put together a mini softbox out of posterboard.  I put that on my speedlight, set it in a little bit to the left and above the figurines in the nativity scene, and then just worked the power level up until I had a proper exposure.  At that point, the only issue left was that some light from the softbox was slightly illuminating the floor and it was reflecting in the globe, so I stuck a leftover sheet of black posterboard on the ground underneath the globe.  At that point, everything was just as you see it in the setup shot, and I made my final exposure.

Post-Processing

The photo above is what I had straight out of the camera, albeit poorly exposure-corrected by my photo manager.  It was pretty much ready to use, save for two issues: the background wasn't perfectly white, and the figurines inside were crooked (I have no idea how they got that way, but they're slanted to the left inside the globe).  The first was a mostly straight-forward fix.  I used a levels adjustment to bring the white-level down a little, and used a white brush to clean up most of the remnants around the edges.  There was still a little bit of gray along the right-most edge of the base (my snoot was a little too narrow, I guess), and of course the bottom had to be removed by hand.  I made a selection around the edges of the base using GIMP's scissor select tool, which I've never used before, and found that it does surprisingly well at following the edges in the image.  I inverted the selection and then took a white brush to the extraneous stuff along the bottom and the right, leaving the base intact.  This left me with some slightly jagged edges, but a little bit of Gaussian blur applied to just the edges fixed that right up.

Then I had to rotate the globe so that the figurines would stand straight up, which was a much bigger problem.  I made two layers and copy/pasted the contents of the globe and the base onto separate layers (using the scissor-select again, with a little hand-editing in quick-mask mode) and then rotated just the globe layer.  This worked surprisingly well---hooray for spheres---and by just dropping the rotated globe down a couple pixels (with the base layer above it, of course) I was able to make it fit almost perfectly.  I just had to layer mask the bottom left corner a little bit to get the edge of the rotated globe and the original globe (still on the background layer) to mesh just right and erase the top of the original globe from the background layer that was sticking up over the top.  With all that done, the rotated globe fit in its place, and everything looked natural as could be.

Except for one final problem.  It occurred to me then, after all that time spent rotating and masking and making everything fit perfectly, that when I'd done the original levels adjustment I'd brought the levels up too much on the figurines, and now a lot of the detail in them was blowing out.  Rather than re-do all that work, I just opened the original RAW conversion up in a new window, lassoed the area I'd messed up, copied it and pasted it into a new layer in the snow globe image.  Rotating the replacement figurines to fit was a little difficult, but I made it work by measuring the angle between two points in both the rotated and unrotated figurines, and then rotating the unrotated ones by the difference.  Then I just moved them into place over the ones that I'd messed up and used a layer mask to make sure they didn't interfere with any of the surroundings.

When I was finally happy with everything, I cropped it down to an 8x10 image and called it quits.  This certainly entailed a lot more post-processing than I'm used to (and the final composition is going to require a lot more), but the end result was definitely worth it.  Tomorrow I should have part II ready, and I'll get to reveal the final composition.

The softbox is a little more than four times the size of the bare flash head.  It's not a whole lot softer than a bare strobe at any significant distance, but in close it does pretty well.  In my case, it's going to be used on a snow globe, so I'll be working with a light source almost as large as my subject at close range.  I got my girlfriend to let me use her as a guinea-pig for some test shots, and in close it makes a half-decent key light for a portrait.  At that kind of range, though, falloff is fast, so it's important to have a good fill light to back you up.

Putting it together is a little time consuming, but not too difficult.  Just grab the template files and print them at 200 DPI, two copies each of the side and top/bottom templates, and one of the panels template.  Super-glue the sides and top/bottom templates to a sheet of black posterboard, cut them out, and cut notches so that you can fold the flaps on the edges down.  Once the flaps are folded down, you'll want to cut the copy paper off of the flaps so that the posterboard pieces stick together well, but leave it on the inside of the softbox.  Glue all four pieces together (a little bit of superglue on the flaps should do the trick) at the base that wraps around the flashhead, and then glue the smaller diffusion panel (I just cut the panels straight out of white copy paper) in place right where the softbox starts to widen up.  You may need to cut the smaller panel down a little bit to make it fit in there: in mine, the flaps of the outer pieces made it a little difficult to fit in.  The smaller panel diffuses the light before it gets to the actual surface of the softbox, to make sure that you get a nice evenly lit surface on the front.  I haven't tested it, but I suspect that without that panel the base of the softbox would act as a snoot, and you'd get a nasty hotspot in the middle of the outer diffusion panel.

With the small diffusion panel in place, just glue the rest of the softbox frame together by the flaps.  Before attaching the outer diffusion panel, you'll need to trim a little bit off the front of the frame where the flaps stick out; sorry, I didn't take the time to properly bevel them in the template.  Then just cut out the larger diffusion panel and glue it in place on the front.  I glued it with the flaps on the outside, but if you're really adventurous you can try to get them glued down on the inside.  Then just let it dry, and your softbox is ready.  It should be just about the right size to stick on the front of an SB-25 or SB-26 with a little bit of velcro on the head, and it should also fit the 580EX, but I haven't tried it on mine yet.  If you have a flash with a smaller head, feel free to modify the templates and repost them.  If it helps, I've uploaded the raw XCF files with the guidelines I used to draw the templates.

I haven't gotten a chance to shoot anything recently, and my screencasting plans have fallen by the wayside thanks to technical problems, so today I decided to dig up some photos from an older shoot and write a post about it.

A little over three years ago, I shot my friend David's senior portraits.  We were both competitive divers at the time (although the "competitive" part was debatable in my case), and he wanted a set of diving portraits, so I went with him to practice one day to shoot.  I had a 300mm f/2.8 lens checked out (one of the perks of editing the school's yearbook), and I planned to put it to good use.  I thought we'd get there a good hour before the sun went down, during which time I'd make some nice available light shots with the 300, and then once the sun set I'd play around a little bit with some lights.

When we got to the aquatics center, we still had a little less than an hour of sunlight left.  That would have worked great, except that that day the coach decided to have David work exercises and drills out of the pool for the first half-hour or so of practice...by the time he got on the boards, the sun was sinking low in the sky.  Even that wouldn't have been too bad, but the area with all the exercise equipment at that center is pretty tall, and it's located directly to the west of the pool.  Not long at all after I started shooting, the sun dipped below the top of the gym building, and I was left with the worst of both worlds: my subject was in the building's shadow, but the sky was still brightly illuminated.  I managed to get some pretty standard, mediocre shots with the 300, but certainly nothing to write home about.

Soon after that, though, the sun went down all the way, the sky went black, and then the real fun started.  I pretty quickly figured out that there was no way I'd get anything reasonable with what little lighting the aquatics center had on the pool (1/50 doesn't work so well when your subject is twisting and flipping while falling), so I broke out my strobes and decided to see what I could do.  I went for a simple cross-lighting scheme, as you can see in the image below.  One of the lights is obvious enough, and the other was located just where the red arrow is pointing.  David was diving from the 1m boards in between the two lights.

I started off shooting from behind the boards, which is where I got the shot at the top of the page.  I had to do a little bit of a levels adjustment in post to get the background to pure black, but it was pretty much there straight out of camera.  The sky had basically gone black, and all that showed up was David and the drops of water flying through the air.  Well, there was a big obnoxious lap clock in the background somewhere, but I cloned that out.

Of course, by this time I'd abandoned the 300 and moved on to my 50mm f/1.8, which worked great at that range.  Autofocus was basically out of the question with the ambient light levels I was working in, so I just set the lens to manual and focused a couple of feet in front of the board.  With my aperture stopped down a little bit I had enough depth of field that I could just point and shoot while David executed his dives.  The cross-lighting gave me a dramatic look that I liked, and I just snapped away like that until I had some good keepers.

While the behind-the-board shots turned out to be my favorites, I did spend some time working from another vantage point before I left.  That roof-looking thing at the top right of the setup shot is the five meter platform.  Normally you dive from it, but I climbed up on top of it to shoot down on David, after securing my camera to the platform with a rope, of course.  It was a little bit of a challenge to keep that large arm structure with all the rope hanging off of it out of my shots, but I managed to shoot around it for the most part.

If I had to choose, I'd pick my ground-level shots over the ones from the five meter platform, but I still liked the shots from above a lot.  If nothing else, it's certainly the only time I've ever gotten to shoot a diver in flight from above.  At the end of the day, I walked away with a nice set of photos worth keeping from my simple little lighting setup, and I got a chance to try something I'd never done before.  To date, that's the only time I've ever shot sports with my own lighting. Things didn't go at all according to plan, but once everything was sorted out I was quite alright with that.

Lately I've been pondering the idea of putting together some simple screencasts about basic post-processing techniques and workflow.  I've received a lot of positive feedback about the idea, and decided to go through with it.  I'll be using a free software workflow (Shotwell, UFRaw, and GIMP) on a GNU/Linux system, but of course the principles will apply to any editing software.  With any luck, I should have the first one together this Wednesday.

IF YOU'RE THINKING ABOUT TRYING THIS, PLEASE HEED THE SAFETY WARNINGS AT THE BOTTOM OF THE POST.

Breaking the Glass

Possibly the most difficult part of the process is getting the glass off of the bulb without messing up the filament too badly.  For my shoot today, I bought a pack of four bulbs; if you're trying this for the first time, I strongly recommend more.  Initially, I tried drilling a hole in the base of the bulb to see if I could let the gas escape with the glass still intact, so it would burn inside the bulb.  Unfortunately, it turns out that the glass is sealed above the metal base, and the bulb with a hole in the base just burned normally.  Since that didn't work, I took the more blunt approach: breaking the glass.

No one wants broken glass everywhere, so before breaking a bulb I stuck it in a pair of old shopping bags.  I just took a pair of pliers and whacked it hard enough to break the bulb.  I'm pretty sure the pressure inside the bulb is significantly lower or higher than the pressure outside it, because it pops and shatters nicely when you hit it hard enough.  The trick to this is that the filament is extremely fragile, and it's exceptionally easy to break it while shattering the bulb.  Out of the four bulbs I shot, three of them came out of the bag with broken filaments (the one that didn't break is what you see in the image above).

Thankfully, a broken filament doesn't mean you can't light the bulb.  The filaments (in these bulbs, at least) are like coiled little springs stretched between two poles, with a kind of hoop sticking up in the middle that it feeds through.  When the filaments came out broken, I just pulled both of the ends over to that hoop and nestled them in so that they stuck to it---think of it like wedging a pencil into the side of a slinky---and then electricity conducted through the hoop, connecting the ends of filament, and the bulbs still burned.

The more filament you end up with intact, the longer it will burn.  The worst one that I pulled out had probably 2mm of filament total, and it burned out so fast that I couldn't even trigger the camera while it was lit.  The longest one, of course, was the one that never broke.  It had easily 3 or 4cm of filament intact, and it burned for nearly two seconds.

The Setup

The photographic setup was pretty straightforward.  Black background, lamp to hold and power the bulb, and a single speedlight to light the holder and the smoke.  Here you can see the whole thing set up on my messy back porch.  For a background, I just taped a sheet of black poster board to my old heavy bag.  I clamped the lamp onto a light stand (and left the shade part hanging on the cord, because it wouldn't come all the way off) with a power strip nearby to switch the power on and off, and put the speedlight back towards the background shooting into the general area the smoke would rise up into.  I put my homemade cardboard gridspot on it to keep the light from spilling over onto the background.

With all that set up, I just put my camera on my crappy little tripod and attached a remote release to it.  I put a bulb in the lamp, switched the power on, and triggered the remote release as soon as I saw it light up.  I started at 1/250, ISO 200, and f/5.6, and it really blew the filament out more than I cared for.  On my final attempt I stopped down to f/11 (and turned the light up to compensate), but that bulb came out with a super short filament, and it burned out before I could get a shot of it, so I had to go with the blown out bulb.

Post-processing

What you see above is the image that I ended up selecting from the bunch, straight out of camera.  The first thing I noticed about it was that the smoke wasn't as bright as I liked, so I made two separate RAW conversions, one to get me the ideal exposure for the smoke, another for the bulb.  I masked those two together, merged the whole thing, and then made a levels adjustment to drop the entire background to pure black.

Next I decided that I wanted the lamp arm on the right side, so I made a rough lasso selection of it, pasted it onto a new layer, flipped it, moved it into place next to the bulb, and just masked it in so it would look convincing.  Then I adjusted the color balance on that layer a little to make the arm blue, so it would look like it was lit by the flash, when in fact it was lit by the burning filament.  Of course, I also erased the original arm from the left side, just using the brush tool with black.  That's one of my favorite parts of working with pure black or white backgrounds: you can remove extraneous bits just by brushing in white or black, no need to clone in the background.

After that, it was just housekeeping.  I erased that purple flare by the filament with the brush tool, and cropped the whole thing to 8x10 to clear a little bit of the empty space on the right side.  After that was all done, I had the final image you see at the top of the page.  All in all, it's a pretty easy shoot to set up, and turns out some nice images.  Just be sure to get your light bulbs while you can.  In 2012, they're going to start phasing out incandescent light bulbs in the US (if you're in the EU, I think you have even less time), and by 2014 they'll be all but impossible to find.

I guess before I close out, I should throw in the standard safety warnings, just in case.  DO NOT TOUCH THE FILAMENT OF A BULB WITH ELECTRICITY RUNNING THROUGH IT.  BE CAREFUL IN GENERAL AROUND ELECTRICITY.  DO NOT ATTEMPT THIS WITH A COMPACT FLUORESCENT BULB: YOU WILL NOT FIND A FILAMENT INSIDE, BUT YOU  WILL FIND DANGEROUS MERCURY.

I got a light stand out, attached an umbrella to it, reached back in my vest for one of my speedlights...and they weren't there.  I'd made the entire journey back to the pond, finally ready to make the portraits I'd always wanted to, and I'd left the lights in the car.  I shot some ambient light photos while we were back there, but there just wasn't much interesting happening, light-wise.  It was a choice between harsh patches of light that occasionally broke through the trees and dull, flat shade (and dark enough shade that I would have been up at ISO 1600 or higher), and none of it turned out worth keeping.  So we decided to head back to the front of the park where I could retrieve my lights from my car, and just use the rock formations at the front of the park.  Ultimately, I think I ended up with better shots than I could have made if I'd gotten to work with the pond like I'd wanted.

I started out shooting with the rock formations and trees just as a backdrop, and set up my lights in the open area out in front of the rocks.  Here's a setup shot from that open area.  It shows two lights, but ultimately I decided only to use the one in the umbrella, and left the backlight out.

It was windy that day, but I have a trick for that.  For one of my light stands (I still need to make one for the other) I have a hand-made strap, just a length of rope tied around some round metal snap clips at either end.  One of the clips snaps perfectly around the stud on the top of the stand, and the other clips onto one of the legs once it's folded up.  If I need to keep the stand up on uneven ground, especially if there's wind, I just let it lean down the incline, leave one end of the strap attached to the top of the stand, and loop the other end around something stable, a small palm tree in this case (you can see the rope in the setup shot stretching out from the top of the stand).  If it's really windy, I'll set the stand right next to something nice and sturdy, wrap the strap around it really tight, and then attach the other clip to one of the tightening pegs on the stand.  Works like a charm, and it's a lot more convenient than carrying around heavy sand bags.

Once the umbrella was up and stable, I just did some straightforward full-body shots balancing the shoot-through umbrella with the ambient light in the background.  I use a set of Radiopopper JrX triggers, and I've modded my Nikon speedlights (I may post about this later) so that I can just connect them with a 3.5mm audio cable and get remote power control.  I worked the umbrella light in to fit just right at my max sync speed with a wide open aperture, to get a little bit of bokeh going on.

After that, we headed over to the rocks, and I had a somewhat more interesting idea.  I decided to try sticking a light inside one of the crevices in the rocks, and getting the model to look into it.  I started out just sticking one light in a crevice, and it made for an interesting effect, but the light was just too harsh.  So I put a stand up on top of the rocks with an umbrella, and pointed it a little bit above the model's head to keep it from spilling down onto the rocks any more than necessary.

Of course, the wind was still a problem, but the top of the rock formation there actually used to be a fountain, and now the concrete is cracked and broken all around the middle of it, leaving just a little bit sticking out from the edges.  I just splayed the legs out nice and wide, stuck them in underneath the concrete edges, and that stand wasn't going anywhere.  With that setup, I got the shot at the top of the page.

To round out the session, I just had her climb up to the top of the rock formation right in front of the umbrella, moved the light that had been in the crevice onto a stand and set it off to the right a little for some fill light.  Then I stuck a 1/2 CTO gel on the umbrella and set my white balance to tungsten, so the background would come out just a little blue in the end.  I shot some headshots and some full body shots, and we called it a day.