When you version one video into many languages, the layout does not stay still, the text does not stay the same length, and something has to give. The only real decision is what gives: the text, the box it sits in, or the timing around it. Teams that decide this once, inside the template, ship clean batches. Teams that decide it row by row, on the day, end up fixing clipped words by hand.
This article is for the motion designer who has been handed a finished English lower third and told that it now needs to run in a dozen languages by the end of the week. It covers how to think about text that grows and shrinks inside a layout you are not allowed to redesign, how to build the template so it absorbs the change, and how to find the rows that still break before you spend a night rendering them.
Decide who gives way before you build anything
A fixed layout is a promise. The brand team signed off on a bar of a certain width, a title of a certain size, and a logo that sits a certain distance from the edge. Translated text does not respect that promise. A phrase that fits on one line in the source language may need two lines in another, or it may leave a third of the bar empty. Neither result is a bug in the translation. Both are properties of language, and your template has to expect them.
So before you open After Effects, sort every piece of text in the design into one of three kinds. This is general practice, not a rule from a manual, but it saves more time than any feature does.
The first kind is text that must stay on one line at one size. A person’s name on a lower third is the usual case, because the style guide says so and because a two line name looks like a mistake. For this text, the box is fixed, the size is fixed, and the only thing that can give is the copy. That means someone has to write a shorter version of the phrase, and it means you need a place in the data to put that shorter version.
The second kind is text that can wrap. A caption, a description, or a call to action often has room to run to a second line. For this text the box is allowed to grow, but only in one direction, and everything near it has to know about it. This is where layout rules earn their place, and we will come back to them.
The third kind is text that can change size within a range. A headline might be allowed to step down a little before it becomes unreadable. Here the text gives way, but only down to a floor that you and the brand owner agree on in advance. Below that floor, you go back to the first kind and ask for shorter copy.
Write this sorting down next to the template, one line per text layer. It sounds like paperwork. It is the document that answers the question a translator or a producer will ask you on the third day, which is “is this allowed to be longer?” Without it, the answer is always “it depends,” and a template with no stated answer gets argued about language by language.
There is one more thing to settle at this stage. Decide whether the source language is the longest language or the shortest. Many teams design around English and then discover the template only works for languages that run shorter. If you can, design around your longest likely case and let the shorter ones sit with some air around them. A bar that looks slightly generous in English is much better than a bar that is cramped in the language that matters most to the client.
Build the template so the layout follows the text
Once you know who gives way, the template has to be built so that giving way happens on its own. The alternative is a template that looks right for the one row you tested and fails on the next.
Start with how the layers relate to each other. Templater applies spatial layout rules when it updates a composition with your data. It adjusts dynamic layers relative to their containing composition and to their sibling layers, so you do not align, scale, or position each one by hand. We describe the feature in more detail in our post on layout rules in Templater. For a multilingual job, the useful idea is that the relationships are the design. You are not specifying that the title sits at a certain coordinate. You are specifying that the background bar stays wrapped around the title with a certain margin, that the subtitle stays below the title, and that the logo stays clear of both.
When those relationships are rules instead of coordinates, a longer title pushes the bar wider and the subtitle moves down with it, and a shorter title pulls the whole group in. The design survives because it was never defined by pixels in the first place.
Then consider your fonts. A single text layer in Templater must use one font, one size, one weight, and one style. That is a stated limit of the product, and it matters here for two reasons. First, you cannot mix a heavy word and a light word inside one layer to solve a spacing problem. If you need that, you stack adjacent text layers and attach them to one another, which our documentation describes as a partial workaround. Second, the font you pick has to carry the characters of every language in the batch. Templater reads your text as it appears in the spreadsheet and accepts characters from any language, but the installed font needs the glyphs. A font that looks lovely in English and has no glyphs for another script will not give you a shorter or longer word. It will give you empty boxes.
This is why font choice is a layout decision, not a styling decision. Two fonts with the same point size can set the same sentence at noticeably different widths. If your brand font does not cover a language you need, you will have a second font in that language’s template whether you wanted one or not, and the widths will change again. Know this before the batch, not during it.
Now think about time. A text change can also be a timing change. A longer phrase takes longer to read, and a title that was on screen for a comfortable moment in the source language may flash past in another. Templater’s Time Sculpting feature lets you apply rules that crop, trim, shift, and stretch layers so a design holds together when the source footage changes length. We should be exact about what it does. It is built for layers of varying duration, which is why it keeps a template intact when footage runs longer or shorter. It is not a feature that measures how long a sentence takes to read, so do not assume it will extend a title for you. Treat reading time as your own design rule. Give text a minimum on-screen time that suits the longest language, and let the shorter ones sit a little longer than they strictly need.
A concrete example makes this less abstract. Picture a sports network that produces a lower third with three fields: a player name, a role, and a short stat line. The English version uses a bar with the name at a large size, the role beneath it in a smaller size, and the stat line to the right in a narrow column. Now the network wants the same graphic for a second and a third market. The name does not change, because names are names, though a name written in a different script is still a different width. The role changes length, because a translated role is a different phrase. The stat line changes the most, because it is a sentence.
The designer’s template treats these three fields as three different kinds of text. The name is fixed size and on one line, and the template simply expects the right script and a font with the right glyphs. The role is allowed to wrap, and the bar is built to grow downward, with the logo anchored to the top edge so it never moves. The stat line has a size floor, and when the text would fall below it, the data source carries a second, shorter version that the template picks up for that row. Nobody opens the project to nudge anything. The sheet decides.
That last part, a shorter version in the data, is a habit worth building. A translated spreadsheet with one column per field will always produce a row that does not fit. A spreadsheet with a “long” column and a “short” column for the fields that matter lets the template choose, or lets a human choose in advance, and the choice lives in data instead of in somebody’s memory. Templater reads a single data source, and it maps data to a single worksheet, so plan the columns in that one sheet with this in mind.
There is one more tool for the cases where one layout cannot flex enough. Templater can switch projects from data, so a row can name which After Effects project template it should use. If one language genuinely needs a taller bar or a different arrangement, you can build that second template and let the row choose it. Be careful with this. Every extra template is another file to maintain and another place for the brand to drift. Use it when the alternative is a layout that fights the language, and not as a way to avoid doing the layout work once.
Find the rows that still break before you render all of them
No template absorbs everything. The goal is to find the rows that fail while it is still cheap, which means before the render queue starts.
The first check is the preview. Templater lets you preview compositions with your data before you render or replicate them, and you can step through the jobs to see how it treats different text, footage, and colors. For a language batch, this is the moment to look at the longest rows and the shortest rows first. Do not walk the list in order. Sort by length of text in the sheet, or simply look at the fields you already know are tight, and go there. If a row is going to overflow, it is almost always one of the longest. A long run of ordinary rows tells you nothing you did not already know.
The second check is spot checks, available in Templater Pro and Bot. A spot check is a still image or a short clip of selected frames from a data-merged composition. You pick the frames by naming composition timeline markers, and if you do not, Templater samples evenly spaced frames across the duration. For a layout problem, the marker is the useful part. Put a named marker at the moment when the title is fully in and the stat line has landed, because that is when overflow shows. Then every row produces a still at exactly that frame, and you can review a few hundred stills in a grid far faster than you could watch a few hundred videos. We cover the mechanics in our post on generating spot checks for quality assurance.
Be honest about what a still can and cannot tell you. It will show a clipped word, a bar that ran past the edge, and a logo that got crowded. It will not show a title that leaves the screen too early, because that is a property of time. For timing, add a second marker near the end of the title’s on-screen life, or render one short clip instead of a still for the languages you suspect.
The third check is the data itself. Many overflow problems are visible in the sheet before you ever touch After Effects. Add a column that holds the character count of each field, or sort by it, and compare against the longest string you tested the template with. This is general practice, and it is crude, but it finds the outliers in seconds. Remember that character count is not width. A line of narrow characters and a line of wide characters can have the same count and very different widths, so treat the count as a way to choose which rows to look at, not as a verdict.
The fourth check is the font dependency problem, which shows up as a failure and not as a layout glitch. After Effects will fail a render item when it cannot find an asset, a footage file, or a font in its expected location. If you are rendering across several machines, every machine needs the fonts the template uses. Our guidance for QUE satellites says the same thing about fonts and plugins: you have to verify that they are installed on any satellite you assign a job to. A language batch that renders fine on your workstation and fails on the farm is often a missing font on a second machine, and it is best to find that out with a small test batch.
When a row fails a check, you have three honest options, and they match the three kinds of text from the start of this article. You can change the copy, which is a request to the translator and a new value in the sheet. You can change the size, within the floor you agreed. Or you can change the layout for that language by routing the row to a second template. What you should not do is open one rendered video and fix it by eye, because the fix will not survive the next time the text is updated.
If you do need to hand finish a version, Templater’s Replicate function is the cleaner way to do it. Instead of rendering, it inserts your data and saves a copy of the project for each version, in a tidy folder structure. You can then adjust the one version that needs a person’s eye and render it from there. Use this for the exception, not for the batch. The point of the whole system is that the sheet and the template decide, and a person only touches what the system could not resolve.
Keep a record of what you changed and why. When the same field fails in the same language a second time, you want to know that the answer was a shorter phrase and not a smaller font, because the next job will have the same problem and you will want to start from the answer.
What this does not solve
We should say plainly what a template cannot do. It cannot tell you whether a translation is good. It cannot tell you whether a shortened phrase still says what the original said. A layout that fits is not the same as a message that lands, and for that you need a person who reads the language.
It also cannot make a bad template flexible. If the design was built with every element pinned to a coordinate, no amount of data discipline will rescue it. The cheapest time to fix that is before the first translation arrives. If you inherit a template like that, rebuild the relationships first, then the language work becomes straightforward.
Finally, there is a limit to how far one layout should be stretched. Past a certain point, a template that has been bent to fit every language fits none of them well. When two languages disagree about what the design needs, a second template that is designed on purpose will look better than a first template that was forced.
The bottom line
Text expansion is not a surprise. It is the most predictable thing that happens when you translate a video, and the template should be built for it from the first day. Sort every text layer by who gives way. Build the relationships between layers as rules, pick a font that covers every script in the batch, carry a short version in the data where space is tight, and look at the longest rows with spot checks before you render the rest. If one layout cannot hold, give that language its own template and say so out loud. That is a design decision, and it is cheaper than a night of fixing clipped words.
We covered the broader picture of versioning video by language in our post on localization with Templater, and the product itself is described on our Templater page. If you are weighing where subtitles fit, our walkthrough of dynamic subtitles in After Effects shows how a second kind of text moves through the same data source. For choosing which data source to use for a language sheet, see our list of tested data sources, and for timing rules that hold when footage changes length, see Time Sculpting.
These are the questions teams ask us most often when they reach this point in a multilingual job.
Frequently Asked Questions
How do I stop translated text from overflowing a fixed bar in After Effects?
Build the relationships between layers as layout rules so the bar follows the text instead of sitting at a fixed width. Templater adjusts dynamic layers relative to their containing composition and their sibling layers, so a longer title widens the bar and moves the subtitle with it. For text that must stay on one line, carry a shorter version in your data source and let the row pick it. Then preview the longest rows before rendering.
Can one text layer use two fonts to fit a long word?
No. A single text layer in Templater must use one font, one size, one weight, and one style. If you need different styling in one line, the workaround is to place adjacent text layers and give each its own attributes using layer attachment properties. It is not ideal, so for multilingual work it is usually better to choose a single font that covers every script in the batch and let the layout rules handle the spacing.
Does Time Sculpting lengthen a title when the translation is longer?
Not by measuring text. Time Sculpting applies time-based rules that crop, trim, shift, and stretch layers so a design holds together when footage runs different lengths. It is built for varying footage duration. If a longer phrase needs more reading time, set that as your own design rule, give titles a minimum on-screen time that suits the longest language, and check it with a spot check clip.
How do I check hundreds of language versions without watching every video?
Use spot checks in Templater Pro or Bot. Name a composition timeline marker at the moment the text is fully on screen, and Templater exports a still image of that frame for every job. You can then review the stills as a grid. Without named markers, Templater samples evenly spaced frames. A still shows clipped or crowded text, but not timing, so add a marker near the end of the title for timing checks.
What if one language needs a different layout altogether?
Templater can switch projects from data, so a row in your spreadsheet can name which After Effects project template to use. Build a second template for the language that needs a taller bar or a different arrangement, and let those rows choose it. Use this sparingly, since each extra template is another file to maintain and another place for the brand to drift.
Can I use several worksheets for different languages?
No. Templater can only map data to a single worksheet, and more generally it can only pull data from a single data source, whether that is a Google Sheet, a JSON file, or a tab-delimited file. Put every language in rows of one sheet, with a column that tells the template which language each row is, and use consistent field columns for long and short versions.
Why does my translated batch fail on a second machine but not mine?
A frequent cause is a missing font or asset. After Effects fails a render item when it cannot find assets, footage, or fonts in the expected location. If you spread jobs across machines, every machine needs the fonts the template uses. Run a small test batch on each machine before you commit to the full run, and keep asset references stored within the project file.
