Toward Intelligent Subtitling
Part 5: Line Breaks

When it comes to line treatment, today’s standard etiquette is quite simple: respect the space constraints, keep linguistic units to one line, and prefer a bottom-heavy shape when possible. This mantra is echoed in classrooms, at conferences and across style guides, and everyone seems rather happy with it.

To my mind, however, there’s a bit more nuance to this facet of subtitling than commonly believed. So, in this fifth part of the Intelligent Subtitling series, I will delve into it by exploring the four groups of considerations in line breaking: linguistic, acoustic, visual and logical.

We all know to divide subtitle text into sense blocks at the highest syntactic nodes: sentences, clauses, phrases, and so on. It is also understood that, to a large extent, this procedure is a matter of compromise, of deciding on what’s more important in each individual case. Indeed:

At times, reading speed and line length limits can throw a spanner in the works:

1. Linguistic

Style guides: 1, 2, 3

On occasion, the syntax itself can pose a conundrum:

Songs and poetry can live a life of their own:

In rare cases, there isn’t much choice at all:

And every once in a blue moon, you will have to go against the norms and break mid-word:

Now, here’s another, less-spoken-about consideration:

A well-placed break can help the audience avoid confusion when it comes to unintentionally ambiguous or easy-to-misread word combinations. For instance, even though the subtitle below can fit on one line, having “on on” together can lead you to believe for a moment that it’s a typo:

So, to prevent a double take, it’s better to split:

Or here is another example:

Here, “make it” means “succeed”. If you split after “delivery” instead, to avoid having a very short second line, it becomes much easier to misinterpret the text’s meaning:

One more thing worth noting here is that awkward breaks can sometimes be fixed by rephrasing or condensing your translation (or transcription). In the following subtitle, even though we are restricted by time and space, you can omit “Palmer” or “it appears” to avoid orphaning that “it” at the end:

This is something skilled subtitlers do on occasion, sparingly, when there’s no better alternative.

2. Acoustic

Pauses in dialogue — ones short enough to not warrant an ellipsis or a new subtitle — will seldom necessitate a line break on their own. What they will do, however, is tip the scales toward one of the available options:

Compare that with this next clip, in which the pause is on the other side of “about”:

Similar situations, different breaks.

Now, the other components of the soundtrack — walla, background music, sound effects, and so on — as well as their properties, such as volume, speed, pitch and change over time, don’t seem to require breaking differently. And the same applies to vocal delivery of dialogue and to abrupt shifts in it — in rate, intonation, pronunciation, word stress, type (normal vs. mouthed), and so on. (I had to watch a lot of stand-up comedy to confirm this!)

3. Visual

Let’s start with subtitle geometry. For as long as I can remember, we’ve been asked to prefer the trapezoid shape, with the first line shorter than the second one — or the rectangular shape, with both lines of roughly equal length, to minimize the text’s intrusion on the image. Now, I myself have a few issues with this “common wisdom”.

First off, this rule comes from studies that are now rather dated. They used older, standard-definition screens, larger font sizes, and data not informed by the current understanding of how our gaze travels within the frame. Today’s hardware can display subtitles in high resolution, making them perfectly legible despite being smaller, and the text itself occupies less of the screen’s real estate, presenting less of an intrusion. Indeed, a recent study by Olivia Gerber-Morón and Agnieska Szarkowska, done with state-of-the-art eye tracking, demonstrated that “the overwhelming majority of participants made their choices [of preference] based on semantic and syntactic units rather than subtitle shape”.

Second, not all media we subtitle strives to deliver a pretty picture. Instructional videos, educational content, corporate training, public announcements, and the like — these don’t require breaks informed by aesthetics. Even in movies, some shots can be simple or plain visually, in which case subtitle readability gets more weight. This is something we forget about when using the “shape-of-subtitle” rule as a blanket generalization.

Third, the overwhelming majority of seasoned subtitlers I’ve talked to agree than linguistic considerations should trump geometric ones most of the time. Our job is to make subs “invisible”, and adding bad breaks only to get the right shape works against this goal, as it worsens the reading process.

Finally, this guideline — in its current form — needs refinement:

What about top-positioned subtitles? Should they also be bottom-heavy?
- No, quite the opposite: top-heavy.
How about vertical videos? Does the shape matter for those?
- Not nearly as much.
Should we always fit subtitles to one line if there’s enough space?
- No, a “short and fat” subtitle can be a better option, depending on the situation.

Despite their importance, these points often get glossed over in style guides.

Moving on.

Another aesthetic element to consider is film composition. Ideally, the shape of your subtitle should not work against the shot’s structure, whether it be its lines:

Prefer the rectangular shape in “horizontal” shots

Framing:

Avoid crossing frames’ borders

Shapes:

Try to not alter the geometric figures (triangle in this case)

As well as focal points, colors, ratios, etc. — all which are brilliantly explained in this video by StudioBinder and described in relation to subtitling in Pablo Romero Fresco’s great book Accessible Filmmaking.

Here are some free-form examples from creators themselves:

1 / 4

Now, it’s quite important to keep in mind that what you see as a subtitler is not necessarily what the viewer will see on their end — both in terms of the differences between the subtitling tool and the viewing device and also between individual settings and parameters of each screen: subtitle font, font size, positioning, resolution, and so on. Unless you have full creative control over your subtitles’ design and presentation, it’s better to not give this group of considerations too much weight.

Next, we have readability and visibility considerations; there are several to take into account. For starters, we don’t want to obscure important elements in the image: faces, objects, bits of on-screen text, and so on. Moving the sub to the top should, of course, be the first line of action, but when that doesn’t really work — e.g. because there’s something to obscure there too — a strategically placed break can help with the issue:

Some end clients require that timed text doesn’t cover any branding

Beyond that, there’s also the question of image contrast. I think it wouldn’t be an exaggeration to say that almost every seasoned subtitler has encountered at least one situation where they wanted to shift the break just to evacuate their text from that annoying, extremely bright spot in the picture. Here’s a real-life example from Guy Ritchie’s Lock, Stock and Two Smoking Barrels:

A single line break could save “your elf” from getting removed from the bar. Poor guy.

When it comes to on-screen texts, there’s a fair bit of variety, but usually people prefer to copy the breaks from the texts if each original line is logically self-contained — even if the whole thing could fit on one line:

When the texts come from different sources, they normally go to different lines:

One exception here is texting and long passages, which usually get the same treatment as normal dialogue:

And, of course, the spatial and temporal limitations can force you to join on-screen texts of different types and from different sources on one line, e.g. via dashes. The Sherlock series is chock-full of these instances:

1 / 3

4. Logical

In rare cases, the creative intent of what’s shown or what’s said will dictate your choice of where to break. It can be the meaning of the dialogue, like in the following scene from Eat Pray Love. The subtitler opted to respect the pause, and now the holy truth comes in two lines instead of one:

It can be plot or narrative logic, like in Heavenly Delusion’s scene about writing an ad for killing “man-eaters”, where the characters themselves note how the meaning changes based on where they split the text:

The line breaks in the subs must be exactly as they are for the scene to make sense

Misunderstood writing on the office door

“We kill // Man-eats” wouldn’t work

Finally, logical line breaks can stem from cinematography, like in the following shot from The Social Network:

Mark is being interrogated by the university committee. He’s under pressure, uncomfortable, boxed in. The shot composition here aims to convey that inner feeling, and to preserve the original creative intent, we must squeeze the subtitle in-between the two people in the foreground, even though the text can fit on one line. Otherwise, the subtitle sprawls free, going against the filmmaker’s idea.

Conclusion

So, to quickly recap:

1. Linguistic

Divide subtitle text at highest syntactic nodes
Compromise where needed due to time/space constraints, syntax, geometry, etc.
Add breaks to help prevent confusion with unintentionally ambiguous or easy-to-misread wording
Rephrase, condense or omit parts of the text to avoid awkward line breaks (but do so sparingly)

2. Acoustic

Use short natural pauses to decide on the best option (for longer pauses, add a new sub or an ellipsis)
Walla, background music, audio effects, and the other components of the soundtrack don’t play a role

3. Visual

[only when you know exactly how the subs will look for the viewer]

Use breaks to:

Respect shot composition: lines, framing, shapes, focal points, ratios, colors, etc.
Avoid obscuring important image elements
Prevent contrast issues

[in any case]

Copy on-screen texts’ breaks if the original lines are logically self-contained, unless the subtitle comprises texting, is part of a long passage, or the time/space limitations force you to join the lines.

4. Logical