This article originally appeared on our site as a seven-part series of featured articles in 2004. It was recently recovered from an archive and is presented here as a complete article for the first time.
Sometimes when I’m working on new fonts I like to have a classic movie in the background. Recently one of the movies I watched was the great old film “Captain Kidd” starring Charles Laughton and Randolph Scott. While watching it with one eye and working with the other, I noticed that the film featured some very stylish piratically evocative title lettering – not to mention some lovely calligraphic maps. So I thought I’d take a stab at recreating the title lettering as a font and even document it in a series of articles showing how a font goes from an idea to final output.
The starting point for this font was a set of screen shots taken from the DVD of “Captain Kidd”. As is the case with a lot of old films the titles are extensive and show up right at the start of the film against a static backdrop. This makes them relatively accessible for screen shots, or in this case, higher resolution shots taken with a digital camera. It also means that a good selection of sample characters is available for reference. The example to the right is one of the best of these shots, because the lettering is particularly large. Many of the others are not as helpful because the print size is smaller, so some of the character details are harder to make out. However, using the best of the images as a starting point I can get enough bits and pieces to construct most of the other characters in high quality, using the lower quality images as a reference for the more unusual character shapes
The next step is to find an already existing font which is at least somewhat similar and use it as a basis for recreating the characters. This will be done in Photoshop, where the rough character forms from the source font will be edited and modified by hand to match the images from the film. In some cases with clearer source material we might actually import the images and trace them, but these source images are so rough that they will almost all have to be redrawn from scratch or developed out of characters from the sample font. For the Captain Kidd font I chose Windlass as a starting font, not because the characters are very similar, but because the basic curves and the relative weights of the thick and thin parts of the letters are similar to the Captain Kidd lettering.
Next I’m going to go character by character to recreate a good approximation of the original source. The first step in this was to make the Windlass characters 30% narrower, which is still too wide for the relatively narrow characters in the source lettering, but as close as I can get without losing the weight of the thicker parts of the letters. The first character I worked on was the C. After making it narrower I redrew the end of the lower curve to match the example from the movie. Then I rotated the cap on the upper part of the curve to match the source. Finally I changed the weight of the curve a bit by thickening the top and thinning the bottom a bit. I also did a bit of tweaking with some of the other lines using a fine-pencil tool. The end result is pretty close to what we’re looking for. On the left you can see the source C and on the right you can see the drawn version. Ultimately I may do a little more work to make them match, like duplicating the notches in the source, but as they stand now the basic shapes are very close.
That’s where I’m stopping for today. Stay tuned for the next installment where I recreate some of the more radically different characters, attempt to import some elements directly from the source graphics and even borrow bits from some other fonts to get the right look.
Having completed the basic planning, the next step is to develop a more complete character set. I started out with the 8 characters (plus variants of the ‘I’ and ‘D’ characters) from the title screen because it has the largest size characters and is the best sample for reference. In the process I had to discard the second ‘A’ because it was too unclear. That screen was clear enough that I can lift the character forms out of it with only minor changes and a bit of clean-up. It was a pleasant surprise to find that I might not even need the ‘C’ I constructed earlier because I got a nice clean copy of the original source ‘C’ to work with. Because these are the cleanest letters I’m going to get, I’ll borrow elements from them to construct some of the other characters.
The next step of this is to make working versions of as many characters as I can working closely from the other screen captures I made. Some of these samples are better than others. I’ll start with the best samples and work my way up to the characters which have to be created entirely from scratch.
The next best screen images I have to work with are the screen with the names of the three principal stars and the screen with the name of the producer. These are good sources because the names include a number of characters which are challenging like B and G. The drawback is that the source characters are smaller and less clear, so they’ll require a lot more clean-up and original work.
When I first saw these screens I had high hopes of being able to lift some character forms almost directly out of them as I did with the title screen. The bad news here is that even though these are the best remaining samples, they just aren’t high enough resolution to draw much from directly. There won’t be any cutting and pasting characters directly. They can be used as a starting point of reference, but they’re just too low resolution and blurry to do much with directly. That means a lot more original design work and much slower overall progress. To get an idea of how blurry the characters from these screens are, take a look at the sample ‘B’ from ‘Britton’. It doesn’t show the same kind of crisp definition you find in the constructed ‘C’ above, or even in the letters from the main title screen. At best it can function as a guideline for characters which will have to be substantially redrawn.
To start off here I take my highest resolution copy of the screen image and strip away everything but the black portion of the main characters. To make this work I have to increase the contrast of the image so I can select the dark portions of the characters relatively clearly. The end result is something like the sample to the right, where I’ve worked with the single name ‘Boggeaus’. I use the magic wand tool to select the dark bodies of the characters, then the ‘select inverse’ command to select and delete everything in the image but the characters I selected with the magic wand. The result is rough, but gives me at least something to work with.
From here the process is to start building the rough source characters up into complete and usable characters, borrowing bits from the small set of 8 good characters that I started with, from the Windlass font which I’m using as a secondary reference, and drawing bits in as needed. You can see a preliminary example of this to the left, where I’ve taken the rough ‘B’ from the source and perfected it using bits of the better quality ‘D’ and ‘P’ characters. The end result is good enough to use.
Ironically I find that explaining how I work these characters into shape is far more tiring than actually doing the work, so that’s where I’m going to stop for now. More characters next time.
This time we’re doing 3 characters, starting in alphabetical order and skipping the ones we’ve already done the easy way. E, F and G are our targets, and they provide good examples of working from blurry examples and using source material from existing characters to produce a good looking result.
Our source E is just terrible, so I decided to make a new E which fits the style of the font, but is a bit more interesting and a lot clearer than our source character. To do this I took the B and the T and used elements (circled in red) from each of them to construct a new E. I used the left half of the B as a starting point, used the right side of the T’s crossbar for the top of the E and the left side (rotated) for the bottom leg of the T. Then I drew in the central arm to make it complete. Once I had the T constructing the F was almost no work. I just took the bottom leg off of the E and added the right hand serif from the bottom of the T. To make the E and the F look less like their sources I added different notches by hand.
The situation with the G was similar, though the source character isn’t quite as bad. To make it I started with the C which I constructed earlier and added parts of the T and I. The parts which I used are circled in red. In this case I started with the left half of the C and then added the right half of the crossbar from the T to make the top right terminus. To fit the curve this element had to be skewed and somewhat rotated to get the right look. The platform-like riser of the G was constructed from the upper part of the I. The main modification was the addition of a little down-pointing serif on the lower right corner.
The process used to create these three characters is pretty typical of what’s required to make most characters. G is actually one of the harder characters. Some of the upcoming ones like H and L are remarkably easy. Looking at what we’ve done so far you ought to be able to project which other characters can be extracted from what we already have. The E contains the L in it. H is just two Is with a crossbar. R derives from the B and K which we already have. Pretty soon we’ll have a full character set. Look for big progress next installment.
We’re getting closer to the end of the project now. This time we’re going to wrap up four easy characters and several more difficult characters as we move on through the alphabet.
H, J, L and O are the easiest characters in the next set. They can be constructed from already completed characters in a manner similar to characters from last installment like T and G. H comes from combining elements of I and F. J comes from parts of G, T and G. L comes straight out of E with little modification. O is the product of combining the left half of the C with itself rotated 180 degrees.
M is one of the most challenging characters to derive from other characters. Ideally we’d want to have an original source character to use as reference, but we don’t have one in our sample screen shots. So to create an M we combine elements from the N and the A. For the left part of the M we use the left half of the N. That’s combined with the right riser from the A and the left descender of the A with some custom modifications to produce an M which fits the style of the N.
Perhaps the biggest challenge for this set of characters was creating the U and the S, because for these characters we have rather awful sources to work from, and they are two of the most difficult characters to do freehand or to derive from other characters. The U is also unusual, because it has a straight right riser, rather than the curved one typical of many designs. For the U we used parts of the G and I for the right riser, and had to basically draw the left riser from scratch to get the right curve, because the curve on the U is significantly different from any of the other curved characters. The S was less of a challenge because the source is in better shape. For the weight and the middle part of the curve we actually used the source directly (cleaned up and darkened), but to make the ends of the curve look decent and fit the style of the other characters we borrowed caps from the C. The end result is a pretty close equivalent of the source.
There are a few challenging characters left. The W and X are the main remaining challenges. Next installment we should be able to get them done, and then move on to the final installment, where we put it all together into an actual font.
Time for the last few letters in the alphabet which we haven’t already done, including a few of the hardest ones
R and Z are pretty easy ones to get out of the way. R consists of the P with a leg added to it which we can take directly from the K which we had an almost perfect source character for. In this case, because the leg on the K looked sort of funky we’ve got two different versions of the R, one with the bottom leg of the K and one with its upper arm flipped to produce an R with a more normal look. Z is also easy. It’s derived from the top of the E, flipped left to right and the bottom of the L, connected together with a straight slanted line.
The key to the remaining characters (VWXY) is the right-slanted element capped by a serif which all of these characters share, along with the heavier left-slanted piece found in the V, W and Y We have to make these elements to form the V and from there we can adapt them to the other characters. For the thin, right-slanted element we take the left leg of the N, flip it over and use the Photoshop skew tool to slant it to the desired angle. This screws up the serif, so we take a new copy of the serif and past it on the top of the now slanted line. The end result is pretty good. For the heavy, left-slanted piece the flourished, curving cross-piece of the N seems appropriate, trimmed out of the N and then skewed slightly to form the proper angle. These two are put together and given a tappered bottom point to form the V. The bottom point goes slightly below the baseline of the other characters, because that seems appropriate based on characteristics of the K and A which both go below the bottom line and the N which goes above the top line of the other characters.
The W is easy to derive from the V. It’s basically just two Vs superimposed on each other and made slightly narrower. The Y is more difficult. It requires the top part of the V, distorted slightly so that the bottom parts meet and combined with the bottom half of the T. The X is the most challenging. The thinner, right-slanted piece is easy to derive from the thin piece of the V. The thicker, left-slanted piece is derived from the I, which has to be skewed to get the right angle, and then has to have a new serif added so that the serif isn’t too badly distorted.
By borrowing parts of other characters to make the new characters it’s relatively easy to make sure that the new characters for which we have no source to refer to end up looking compatible with the font as a whole. Our next step in developing the Captain Kidd font is to import the characters into Fontographer and put them in their final form. If you look at the full character list below and compare it with the samples we started with, you’ll notice that in the main title the characters are at different up and down positions relative to each other, as well as different sizes. But in making our basic character set we’ve made all the characters the same size and given them the same baseline position. Once we get the characters into Fontographer we’re going to have to change that and reintroduce the size and position variations and do it in a way which looks interesting and not just goofy. This may be our greatest remaining challenge.
Now we’re ready to import the characters into Fontographer and start making them into an actual font.
Right now the characters are in a Photoshop file. To get them into Fontographer we copy and paste them in a few characters at a time. Fontographer doesn’t allow direct conversion of graphics to outline format (all font characters are made with outlines), so you have to trace the graphics in Fontographer. Initially we do this with the AutoTrace feature, which does a semi-adequate job of tracing the outline of the characters. But because these characters have a rough look, the AutoTrace function doesn’t give a really accurate result, so once the initial tracing is done we have to adjust the outlines by hand to match the graphics more closely.
We do this tracing a few characters at a time, because Fontographer’s AutoTrace feature runs into memory problems with outlines which are too complex. Once the tracing and adjusting of the outlines is done we copy the characters out into their individual positions. At this point we can take our first look at the basic character set (shown in the next graphic).
Right away we spot a problem that we didn’t notice as much with the original graphics. The characters just aren’t the same size. This isn’t surprising, since we were working a source where characters were slightly different sizes, but we have to fix it. Ironically, the final font may have some variations in character size, but before we reintroduce those variations we want to have a standard, uniform version of the character set. To regularize the characters we use the Transform feature in Fontographer to do a percentage increase or reduction in size of the characters until the sizes match pretty well. While we’re at it we also do an initial adjustment of the spacing between the characters so they fit together well. The result is shown below.
There are still some problems. A few of the characters (K, X, Z) seem a little under weight, and there are more positioning issues to look at. Those things are easy to tweak. Our next major task is working out final character sizes and positions for the upper and lower case. The upper case will essentially be the basic character set, but for the lower case positions (we don’t have an actual lower case in this font) we will be varying sizes and positions to create the uneven or staggered look of the positioning of some of the characters in the samples from the movie.
Now that we’ve got the basic characters into Fontographer it’s time to play around with them and produce the variations of character positioning which are characteristic of the original lettering. Ultimately we’re going to end up with at least 3 different versions of each character, derived from the original set.
The first variation will go in the lower case character positions and is a standard small-caps version of the letter set. These are made by shrinking the original characters about 20% and adjusting their weight up about 20% so that their thickness continues to match the originals. This is done with the character baseline setting preserved so that the bottoms of the small caps match up with the originals.
The next set of variant characters are the offset characters. These are created by taking the primary characters and moving them about 20% farther down, across the baseline so that their tops will dovetail under the serifs of the standard characters and more or less match the tops of the small caps. These new characters will end up located in the alternate character positions. On the Macintosh these are accessed with the combination of the option key and the character desired. An example of the three different versions is shown with the 3 examples of A to the left.
Finally, there is a fourth version of some of the characters. These are the variant characters, which are based on the standard set but with some changes to the actual shape of the character, such as an elongated tail, a different version of the serif, an extra long ascender or descender, or other changes which vary the way the character looks. Some of these are taken directly from the original lettering and others have been added which are consistent with the style. An example of a character with four different versions is shown with the 4 examples of H to the left.
There is still some work to do. The font still needs numbers and punctuation, plus eventually customized spacing and kerning, so there’s something left to cover next time.
Once we’ve got the basic character forms all sized right and in the proper positions in the font, we need to develop some numbers and punctuation.
We have quotation marks and apostrophes from the original sample. They can also be adapted to form a comma. A period is easy to draw with a rough circle of appropriate size. For the rest of the punctuation we will draw on the somewhat similar Windlass font, import the characters, resize them to fit the shape of the Captain Kidd characters and in this case slightly increase the weighting with a tool which does that automatically in Fontographer.
The numbers require some more direct attention and will be derived from some of the other chatacters. To start with we copy the following letters into the positions of numbers which have some similarity 0=O, 1=I, 2=S, 3=B, 4=M, 5=F, 6=C, 7=Z, 8=B, 9-C. From there we modify and augment the letter forms to create the numbers we need. 0 requires almost no modification of the O character, except making it slightly narrower. For 1 we take the top of the I and slant it and reshape it to form the traditional slanted top of the 1. For 2 we flip the S horizontally, cut off the bottom, import the bottom arm of the Z and link the curved S top with the flat Z bottom. For 3 we remove the left side of the B and cap the top and bottom arms with the top and bottom caps from the S. We cap the middle arm with a created cap similar to the middle arm of the F. 4 is one of the hardest numbers to create. The M has some similar lines, but without a lot of modification it’s hard to make a 4 look just right. We start with the right riser of the M, but want to combine it with the left-hand slanted element and then add a cross bar. To do this we take the M apart and sort of reassemble it to get the right look. Ultimately we have to actually rotate the slanted bar about 15 degrees and do a lot of tweaking to get the right look. 5 is one of the easiest numbers. It’s just the top of the F and the bottom of the S combined. the 6 is the C with the lower curve closed into an appropriate-looking loop. 7 is easily derived from the Z with a broader base and a variation of the pediment from the I added. 8 is one of the hardest numbers to get looking just right. To start out with we take the right half of the B and clone it and flip it and attach the two sets of double half-loops together. This forms the general shape of an 8. The catch is that no 8 is ever really perfectly symetrical, so we have to adjust the weighting by hand to simulate pen-drawn weight variations. This means narrowing where the stroke would be thinner and thickening where it would be thicker. This pen-like weighting is not a pronounced characteristic of other letters like B, P and O, so we only do it a little bit. This is a likely sign that the original lettering was drawn as outlines and filled in like typical advertising lettering, rather than drawn freehand with a pen. The final number is 9, and it’s an easy one – just a flipped version of what we did with the 6.
With that, the numbers are done and we have a complete font. It doesn’t have a lower case, and constructing one to match from scratch would be painfully difficult. However it does have small-caps and a nice set of variant capital letters, and that’s true to the original source lettering. Despite this, bowing to popular demand, we did eventually make a lower case variant version of the font. To do this we imported the lower case letters from Buccanneer, made them thinner, lighter and slightly shorter, and changed their serifs to match the serifs on the Captain Kidd upper case. It’s not entirely true to the original design, but it looks right, and offers another option for users who want lower case letters.
To order the complete set of Captain Kidd visit our ONLINE STORE.