Index Cleaning Scans Typesetting Saving Files Advanced Stuff

2. Typesetting

2.1) Text Content
Use text rectangles. Even if it's just one word to be put down, use a rectangle; don't just click once with the text tool. Rectangles make resizing and re-wordwrapping text infinitely easier, avoids awkward text shapes and the problem with leading and trailing spaces distorting alignment, and allows you to quickly compare different wordwrap schemes to see which is most attractive and fits the bubble best (take advantage of this by trying multiple rectangle widths for text blurbs, even if you think your original looks good, since you may find better ones). Just don't hesitate to put in the manual returns when needed.

Some things to keep in mind:
Don't rely on Photoshop word wrap. Sometimes, based on the text box dimensions, Photoshop will word-wrap right in the middle of a word, and it's better to just manually return and put it on a line of its own. Text rectangles are better because it can word wrap, but still gives you the freedom to insert manual returns as needed. Try to match the general shape of the bubble with your text. Avoid hourglass-shaped text blurbs, since this looks odd (usually you want a convex shape, though sometimes a square/rectangle is more fitting). You may need to use manual returns to get the shape you need. Ideally you should have as few words as possible that occupy more than one line, and never more than two. If a word is too long to fit on 2 lines, adjust the font. One somewhat exception is honorifics (-san, -sama); a name that occupies 2 lines, it's generally alright if you have to put the honorific on a 3rd. Another point about wordwrapping, if there is an ellipse (...) in the middle of a bubble rather than at the beginning or end, try to wordwrap the bubble so the ellipse is at the end of a row; insert a manual return if necessary. If you really can't without the bubble looking awkward, put 1 space after the ellipse before continuing the bubble.

A special case is when photoshop breaks words in half. It can help, as far as centering goes, to manually put in the hyphen (and a space) rather than let PS do it automatically; e.g. if PS breaks the word 'football' in half, if it breaks it after the foot, you'll get 'foot-//ball'. If you replace 'football' with 'foot- ball', you'll notice that 'foot-' is now centered a bit better.

With some fonts (including the common wildwords), for the most part majuscules and miniscules are nearly identical, but there is one more noticeable difference - the letter i. Different groups have different opinions on this, and I personally don't care much since it can still be differentiated from other similar characters, but admittedly at smaller font sizes, using the lowercase i can be a bit harder to see, since it tends to scrunch up against neighboring letters.


A comparison of majuscule and miniscule "i." Using the font wildwords, the first row is typed in all caps: "I LIKE 1." The second row is all in lowercase: "i like 1." The bottom row is what would be proper capitalization, "I like 1."

Use vertically oriented text sparingly. Oftentimes it's better to have a mostly empty but with horizontal text than big tacky vertical text. Also, spacing can be skewed with vertical text; multiple spaces may be necessary between words to be sufficiently clear.

Different groups (and different translators) have different opinions on SFX. Some people simply don’t bother translating them unless they affect the story. Some translate them, but don’t replace them, choosing to footnote everything. Some say translate and replace everything. My personal opinion is to do as much as you can, but if you can’t replace it, then footnote it, and obviously if your translator didn’t translate it you can either try to erase the SFX or just leave it.

Sometimes you'll get SFX translations in romanji, which sometimes isn't very intuitive for english audiences. For situations like these, you can consult lists like the one here to figure out what might work in its place.

2.2) Font Choice
If you use an exotic font, try to fit it to the mood of the source text. Some fonts only look good at small point sizes, and some only look good at large point sizes, take this into consideration. Also, don't make your text too small so that it's hard to read (you should be able to tell how small is too small, take into consideration some people view manga with fit image zoom which can often be in the range of 60% zoom), and don't make it too large that it completely fills the bubble (there should be a decent bit of white space between the edge of your text and the outlines of the bubble, and text should never touch bubble edges). When considering point size, use whatever point size best fits the text and the bubble. If a slightly smaller or larger size would fit better, use it. Don't be lazy and try to get away with one or two point sizes per page, if using other sizes fits better (on the other hand, don't go overboard, otherwise your page will feel inconsistent). Sometimes you may even need multiple fonts or font sizes in the same blurb of text (most commonly italicizing for emphasis or changing font and font size for shouting), do so when appropriate.

Most groups will have their own guidelines for font usage, and if so, follow them. If not, a few things to keep in mind: do not use wildwords for everything (or any other single font). Some scanslators intentionally avoid wildwords, protesting its overuse. While it most definitely is overused, it still is a good font (I’ve yet to see a better one, even among alternatives like animeace, astrocity, mighty zeo, zud juice, etc.), and has become as synonymous to scanslation as Times New Roman is to schoolwork. Intentionally avoiding a good font because others misuse it is dumb (those who boycott it almost always end up choosing an uglier or nearly identical font, but claim some sort of moral victory by having avoided wildwords). However, do make sure to put a little variety in your typeset.

Generally for manga, an all-caps font is used as the main dialogue font (some have started using fonts that aren't all-caps). As most of us are well aware, all caps can oftentimes be abrasive or difficult to read. After all, study after study has shown that one of the ways we recognize words is from the coastline (the "outline" of the word, i.e. x-height, ascenders & descenders, etc). A good article if you care can be found here; lesson 1 sections 2 & 6 cover some of what I just mentioned. So why use all caps fonts if they're harder to read? Because we associate such fonts with manga and comics. Fonts get a lot of their meaning and context from their use in other contexts, especially trademarks or period type, and this associational meaning is actually extremely powerful (think your stereotypical Western or computer font). Of course, it's up to you what you want to use. Personally I say stick with the all caps fonts. It allows you to switch to a more standard non-caps font with more effect and impact (switching from a conventional font to an all caps font just doesn't work as well; italics or bolditalics can cover the 'shouting', but if you're already using a standard font, it's a bit harder to go more subtle).

As for when to use serif or sans serif fonts, the old rule was serif for body text to help define the coastline and help move your eyes along the text, while sans serif for headers or titles or whatnot. This doesn't really hold true for scanslations, in large part because serif fonts can oftentimes look bad on computer screens (especially at smaller sizes). Still, it's not a clear distinction, and even among serif and sans serif fonts, there can be a lot of differences in legibility and meaning. Just go with your gut feeling, consider what moods or emotions come with different fonts, and choose appropriately.

Somewhat common but poor font choices: Having lots of fonts is entertaining, but having too many can be more distracting than helpful. Everyone has their own sense of aesthetics and style, so just do what you think looks good.

A few nice font links:
2.3) Font Placement
Just about all dialogue and thought text should be center aligned (and many signs and whatnot). Even if the bubble is up against a corner, center the text (only use a different alignment if somehow center alignment causes text to contact bubble edges, perhaps in the case of overlapping but distinct bubbles). The only text that should ever contact boundaries (bubble or frame bounds) are SFXs and the occasional small text, and in those cases, try to put a decent amount of crossover to show that it is supposed to obscure the boundary. Take care to center each block of text in its bubble (a few pixels in one direction does make a difference). Sometimes there will be multiple connected bubbles, and their proximity combined with text size and length makes it impossible to center each blurb in its respective bubble. Do the best you can though, just make sure it's legible and that people can distinguish the bubbles (don't line up the text bubbles or else people will get confused).


Take those extra few seconds to center text properly. A few pixels here and there does matter. Also, note how, despite the bubble being up against the corner, the text is center aligned, not right aligned. In this case the bubble is still fairly round; however, even when the edge or corner is a more dominant factor in bubble shape, keep center alignment.

If two bubbles are connected, do not combine the text in them into one blurb; leave them as two, and try to respect an imaginary boundary that would divide the two when typesetting each bubble (sometimes, due to word wrapping issues, you’ll need to cross this imaginary boundary; if it looks better that way, go for it). Sometimes a bubble's outline will appear to be two joined bubbles but actually just be one (in which case, keep it as one text blurb), so just pay attention to how the text was in the original language. On occasion, when the separation between the bubbles is sufficiently minor, one of the bubbles is clearly vestigial (partial bubbles against frame edges, no text in that partial bubble in the raw), or when there are only one or two words of text combined for two bubbles, it may be alright to merge text blurbs.


Note how, on the left, text placement on connected bubbles respects the imaginary boundary between them (dotted line, don’t actually draw a line or anything though, just eyeball it). In the middle figure, as noted, longer words may require crossing bubble boundaries (word-wrapping here is undesirable), but effort is still taken to show that there are two separate text blurbs here. The third example illustrates what not to do.


As a rule of thumb, if, in the raw, the text was split into distinct blurbs, you should do so as well.

Text can also be “framed” behind art, most usually for SFX but sometimes for signs or notes or the like. There are a couple ways to do this: the first and less effective way is to simply rasterize the text, then erase what areas of text the art should show through. The better way is to make a duplicate of the underlying art, place it above the text, then erase the art where the text should show through. Sometimes the magic wand may be helpful in erasing the appropriate sections of the top art layer, sometimes you will need to manually erase. This method is superior to the former in that it allows you to preserve the text, allowing you to change the text if needed, or adjust its location.


Framing text behind art. The original is seen in (A). Erase the original text, in this case using the clone tool, and taking care to align the dots in the background texture (B). Place your text as you normally would (C). Then, going back to the background layer, copy the area of art around the text, and paste it to a new layer (D). Generally, when you paste, the new art layer will still be under the text, so select your new layer and drag it above the text layer in your layers window (E). The text should now be hidden, and the new layer’s art should be aligned with the underlying art properly (if not, adjust the transparency on the top art layer to see what’s underneath, and align it to the background art, making sure to match things down to the pixel). Then, simply erase the areas that the text should show through, while avoiding the areas that you want to cover the text with (F). The final product is shown in (G).

Another method to achieve a similar effect is to place an unlinked layer mask on the text layer. In some respects this may be easier, but a problem arises with layer masks when your text layer has a layer style (i.e. stroke). Although the text is still text, as you place the mask to allow the underlying art to show through, the layer style treats the new border as the text layer edge, so you get unwanted effects against mask edges. If for some reason you want something like this (for framing text behind art you pretty much never do), then by all means.


Using layer mask has detrimental results. Note how the section within her hair (circled) should be plain white, but because of the stroke layer, now has a border (no text should be left in her head anyways, though, so this isn’t a very realistic example). More importantly, as marked by arrows, regions that should simply terminate against her head now terminate with a stroke border.

When placing text on art, it is not necessary to put the text in the exact original spot. For one thing, raw text is generally vertical, while english is horizontal, oftentimes meaning you'll have to choose what to obscure with your text. Assuming you did a good enough job redrawing, if there happens to be a nice open space nearby, or if an alternate positioning is simply aesthetically superior, go with it.

2.4) Distorted Text
There are a few ways to distort text. The first is the warp text ability, which gives you a choice of several preset distortions to choose from. Each of these distortions can be further customized by adjusting the various settings under each, which may dictate how much or in what direction the text is distorted.


After you type your text blurb, with the text tool still selected, click the warp text button. This brings up a drop menu with a variety of preset distortions.

Another way is to free transform the text blurb with CTRL+T, which will allow you to stretch or squeeze text. Also, under Edit>Transform, various other transformations can be applied, some only after the text has been rasterized. Edit>Transform>Distort is one of these, and unfortunately is one of the more useful of the transformations.

2.5) Other Points
There is one main text layer style you should be aware of (that I’ve been mentioning throughout) and that's stroke. Outside, normal blending, 100% opacity, ~2-3 pixel (depending on the relative text size, font choice, and background) white stroke is essentially the standard for putting text on art. If you want, black stroke is acceptable if you decide to use white text, and this helps add some variety to your typeset. No gray stroke (a better option if black on white or vice versa is too sharp is simply white stroke on gray text; grey stroke is almost never attractive, though). Other layer styles are handy, such as outer glow, drop shadow, etc., so use them when appropriate, but stroke is by far the most common style used. Against some textured backgrounds, however, stroke may be difficult to see without using a very thick and ugly stroke; in this case, try a thicker outer glow (the fading in the glow offsets the extra thickness) or perhaps add a drop shadow. Using the occasional gradient or pattern overlay helps add variety as well, and custom patterns can be made by selecting the desired pattern with the rectangle marquee, then going to Edit>Define Pattern (this is also how you use custom patterns for pattern fill).


Stroke (in this case, 2 pixel white) is perhaps the most commonly used layer style for editing.

Typeset is not simply putting down translations; good typeset should help the reader identify various points, such as the mood of the speaker as well as, in some cases, the speaker's identity. Sometimes using a separate font for a certain style of conversation or speaker is helpful, or rotating a block of text can indicate that a speaker outside of the frame is speaking. For SFX, small text, etc., use a different font to help identify this. SFX are almost never level and parallel. Usually, SFX need to be rotated, and if a frame contains multiple identical SFX, they should be rotated to different angles depending on their position in the frame. As mentioned before (but worth mentioning again), never use the same font for everything. At the very least, separate fonts (or font variations, as in the suggested font list above) should be used for dialogue, thought, shouting, and SFX.

Try to use round numbers for main text font sizes (integers or, if necessary, increments of 0.5). Use smooth anti-aliasing. Avoid faux bold or faux italics (they look tacky, and are generally unintended with a given font- if the creators wanted an italics or bold variant, they'd include it; nice site about why faux italics sucks here). Never flatten text onto art. If you used multiple art layers during reconstructing, and assuming you don’t need the extras for framing/layer order effects, you may want to merge those layers to save a bit of processing power. If you need to distort text, then rasterizing the individual layer is fine (though you won’t be able to change textual content), but never flatten it to the underlying art.

Avoid compressing fonts for the sake of fitting bubbles. If you need a little bit more room, it might be all right, but if you have to go to less than 80% in any one dimension (usually width), just don't. Many people erroneously think that a font is only as legible as its largest dimension--they're wrong. If anything, it becomes even less legible than a normal font in the shorter dimension, since the height:width ratio is now distorted. If you've ever had a class where, on a Powerpoint presentation, the professor takes a wide but short table or graph and stretches it vertically to take up most of the slide, you know what I'm referring to.


The top is a sample of text, font Wildwords Roman, size 24pt. Let's say that was too big for the bubble you wanted. The bad way is to just comress the width, in this case, 50%. The text now has the width of a 12pt font, but the height of a 24pt font, and is hard to read. A better way is to resize the font, and re-wordwrap as needed, as indicated on the bottom right. Note that even though the text is smaller in terms of height, it's much easier to read than the condensed text. Not shown is to simply resize to 12pt and keep the wordwrapping the same, but even that is more legible than the bottom left example.

If, for some reason, you were editing from indexed png’s or gif's, you’ll notice as soon as you put down a text box, the rest of the page turns pink (masked). Go back and grayscale (or RGB, but generally if you’re dealing with png’s for manga they’re grayscale and color pages are jpg) the image, then proceed with editing. If you made any changes to the png, be it resizing, erasing, drawing, rotating, or the like, you’ll have to revert to before you made those changes, grayscale the image, and make those edits again. Art (and flattened text) becomes heavily distorted if any alterations are made in indexed mode (if you try to erase or draw, you’ll notice, rather than a smooth circle, your brush will be very jagged and blocky looking).


Index Cleaning Scans Typesetting Saving Files Advanced Stuff