A Templater Bot that runs for hours can get slower as it goes, and our documentation points to After Effects caching as the cause. Turn off layer caching, turn off disk caching on the machine, and the Bot stops carrying weight it does not need.
That is the short version, and it comes straight from our documentation. The longer version is worth reading if you leave a Bot running overnight or over a weekend, because the fix is a few preference changes and the slowdown builds up over a long run. This article explains why the slowdown happens, the exact preferences to change, the Bot settings that control how hard it works, and the project habits that make every job cheaper. It also says what we cannot promise, which is a specific speedup number.
Why a Bot slows down, and the two caches to switch off
The Templater Bot is a mode. When you enable it, Templater checks your data source at regular intervals, picks up rows marked ready, versions the target composition for each one, and outputs the result. It then goes back to waiting. A Bot can run like that for a very long time.
Our documentation describes what happens over that time. If the Bot stays enabled for an extended period, Templater’s versioning process might slow down as it keeps receiving more and more data. The reason is the way After Effects caches data when layer properties change. Every time Templater changes a property to put new values into a layer, After Effects remembers the old state so you could step back.
That memory is a gift when a person is working by hand. You change a color, you do not like it, you undo, and it comes back instantly. A Bot never undoes anything. It only moves forward. So the cache that helps a designer is pure overhead for a machine that does the same kind of change thousands of times. The fix is to switch off the features whose whole purpose is restoring changes.
Turn off layer caching
This setting is not visible in the normal Preferences dialog, so the steps are specific. Start After Effects. Open the Preferences dialog while holding down the Shift key. On Windows, go to Edit > Preferences > General while holding Shift. On a Mac, go to the application menu, then Preferences > General, again holding Shift.
The Shift key matters. Holding it reveals an extra category at the bottom of the list of preference categories. That category is named Secret. Click it.
Make sure the checkbox next to Disable Layer Cache is checked. Then find the field labeled Purge Every ___ Frames During Make Movie and set the number of frames to 0. Click OK.
If you open the Preferences dialog without Shift, you will not see the Secret category and you will wonder whether your version has it. Close the dialog and open it again with Shift held down.
Turn off disk caching
Disk caching has the same shape as layer caching. It lets you restore changes quickly when you work manually, and it does little for a Bot. We recommend that you disable disk caching on any machine where the Templater Bot is left enabled.
There is a version detail here that catches people. Before After Effects 24.3, the Templater advanced preference named Purge all memory caches after each job cleared the disk cache behind the scenes. After a change in 24.3, that preference no longer empties the disk cache after each processed job. It still clears out all RAM, image, and undo caches. So if you upgraded After Effects and your disk cache began growing, this change may be the reason. The preference just does less than it used to.
Follow these steps. In After Effects, go to Edit > Preferences > Media & Disk Cache. In the Disk Cache section, click Empty Cache. If the button is disabled, the cache is already empty. Click OK.
Then decide about the Enable Disk Cache checkbox. If you use this machine primarily or only for automation, clear the checkbox. It works in tandem with the Templater preference Purge all memory caches after each job, so between the two, the Bot has very little to hold on to. If you also design on this machine, or rely on the disk cache for other reasons, skip the checkbox. In that case you have to clear the disk cache by hand after each job run by the Bot.
That last sentence is the honest trade-off. A machine that is both your workstation and your Bot pays for it, either in cache you clear yourself or in a designer’s speed. The cleanest setup is a machine that only runs the Bot. We know that costs hardware, and it is not always possible. If you cannot dedicate a machine, put a reminder next to the Bot’s start button.
What we cannot tell you
We do not publish a figure for how much faster a Bot runs after these changes, and we are not going to make one up. The documentation says versioning might slow down over a long run and tells you which settings to change. How much you gain depends on your project, your footage, how long the Bot runs, and the machine. Measure it yourself. Note how long the Bot takes per row after several hours before you change anything, make the changes, and compare.
Bot settings that control how hard it works
Caching is half the story. The other half is how often the Bot looks for work and how much it takes on at once. These live in the Templater panel under Preferences > Bot.
What the Bot does with each job
First, choose what the Bot should do automatically. The options are Render, Replicate, and Spot Check, and you can tick one or more. Tick the actions your workflow needs and leave the others alone, so the Bot is not doing anything you did not ask for.
Naming Bots
If you run more than one Bot against the same data source, give each one a Bot Name. If you run one Bot per data source, you do not need to name it and your data source does not need a bot column. The data setup for one Bot and for many Bots has its own article in this series. For how a data source carries instructions into a render, see using flags in data sources.
Polling frequency
The field reads Check every ___ minutes for new data, and it exists separately for Google Sheets and for JSON sources. The default is 0.5 minutes, which is every thirty seconds. The minimum is 0.3 minutes, or eighteen seconds, for Google Sheets, and 0.2 minutes, or twelve seconds, for JSON sources.
It is tempting to set this as low as it goes. Do not do it without a reason. Our documentation warns that setting the value too low may cause After Effects to behave unexpectedly. A Bot that checks every thirty seconds is already fast enough for nearly every workflow where a person adds rows by hand, and a faster poll buys you seconds, not minutes.
The queue limit
The field reads Allow up to ___ rows to be queued at one time. This caps how many rows or JSON objects the Bot queues and processes at any given moment. The default is 20, and 20 is also the maximum.
The documentation gives a clear use for this. Say you have a data source with 10,000 spreadsheet rows or JSON objects. You can mark all of them ready by setting the render-status value, then leave the queue limit at 20. The Bot works through them in batches of 20. This reduces the memory Templater needs to process all 10,000.
So the limit is both a protection and a pacing tool. It is the reason you can paste a very large job into a sheet and walk away.
Scripts that run on events
You can register shell scripts and ExtendScripts to Templater events from the same Bot preferences, using Setup Shell Commands and Setup ExtendScripts. They are useful, and they are also work. A script on a job event runs for every job, so a slow script is multiplied by every row. By default, Templater runs shell commands synchronously, which makes After Effects wait until the script finishes. If a script does not need to finish before the next step, tick Asynchronous to run it as a detached process.
One more setting is worth knowing for design sessions. By default, Templater runs your registered event scripts each time you perform a Preview. There is a checkbox, Prevent event script execution on Preview, that turns that off. Use it if previewing in the middle of a design session triggers a script you only want during real jobs. The Templater help pages cover event scripts in detail (see Event Scripts at support.dataclay.com/templater).
One weekend run, set up properly
Here is how the pieces fit together for a long unattended run. A team has a machine used only for automation. On Friday afternoon they open After Effects with Shift held, disable layer caching, set the purge frames to 0, empty the disk cache, and clear Enable Disk Cache. In the Bot preferences they keep the default polling, leave the queue limit at 20, and tick only the actions they need. They set render-status to ready on a very large sheet, enable the Bot, and leave. The Bot takes rows 20 at a time, and the caches that build up over a long run are switched off. On Monday they check the output and the time per row. That is the whole recipe, and every step in it comes from the documentation.
Make the project itself cheaper to render
After Effects is not known for blazing export speeds, especially with complex compositions. We said as much in our own piece on speeding up After Effects exports. Caching settings keep a Bot from getting slower. They do not make a heavy project light. For that you work on the project.
Purge what you do not use
Templater does not decide which layers matter. It evaluates every layer in the project. A layer you forgot about, an effect you left on, or an old composition hiding in the project panel all cost time on every job. Go through the compositions carefully and delete anything unused. It is the simplest tip, and the savings repeat on every row of the batch.
Pre-compose and crop
If a layer is scaled down a long way, or only a small part of it is visible, pre-compose it and crop the pre-composition down to the size you actually need. That cuts the number of pixels After Effects has to process. A full-size source scaled to a thumbnail is a good example. After Effects still handles the large version unless you make the composition smaller.
Tidy the data, use proxies, pick sensible settings
The data source matters too. Templater works with spreadsheets, JSON, and live feeds, and a source that is properly formatted is easier to process. Our article on handling dirty or incomplete data shows what goes wrong when it is not. For heavy footage, use proxies so After Effects is not straining during preview. Choose export settings that fit where the video will be used, rather than the highest quality by habit.
Let the template adapt its own timing
Templater’s Time Sculpting feature lets layers have variable time durations, so timing can adjust to the length of the content in each job. We describe it in time sculpting with Dataclay Templater. It is not a speed feature, but it is a way to let one template cope with different content lengths.
If one machine is not enough
Sometimes a single Bot on a single machine is the limit. Projector, our rendering and distribution app, can use the After Effects rendering engine in parallel based on the number of cores in your machine. You can also run more than one Bot against the same data source by giving each a Bot Name. To plan how much capacity a deadline calls for, read how to size a render farm for unpredictable, deadline-sensitive volume.
A practical order of work, and the bottom line
Do these in order, because the early ones are cheap and the later ones take thought.
Start with the two cache settings on the Bot machine. They are quick to change and address the slowdown our documentation describes. Then review the Bot preferences. Keep the polling at the default unless you have a real reason, keep the queue limit at 20 for large sources, and tick only the actions you need. Then spend an hour on the project, removing unused layers and cropping oversized pre-compositions. Finally, run your normal batch and watch the per-row time. Write the number down before and after each change, one change at a time, so you know which setting earned its keep. If you change the cache settings, the polling, and the project all in one afternoon, you will see a result and have no idea why. If it is still too slow, you have a real capacity question, and more machines or Projector may be the answer. A failing long run is its own subject, covered in what happens when a Templater batch fails mid-queue.
The bottom line is that a Bot slows down because After Effects keeps remembering changes it will never be asked to undo. Disable layer caching and disk caching on a machine dedicated to the Bot, keep the polling and queue settings sensible, and cut the unused weight out of your project. Do not expect a promised number from us. Expect a Bot that carries less cache weight over a long run, and measure it to be sure.
That is the checklist. These are the questions people ask us before they leave a Bot running.
Frequently Asked Questions
Why does my Templater Bot get slower the longer it runs?
Our documentation says that with the Bot enabled for an extended period, Templater’s versioning can slow as it keeps receiving more data. The cause is how After Effects caches data when layer properties change. The fix is to disable layer caching and disk caching on the machine, since both exist to restore changes you never make with a Bot.
Where is the Disable Layer Cache setting in After Effects?
Open Preferences while holding the Shift key. On Windows, use Edit > Preferences > General. On a Mac, use the application menu, then Preferences > General. Click the Secret category at the bottom of the list, make sure Disable Layer Cache is checked, and set Purge Every ___ Frames During Make Movie to 0. Without Shift held, the Secret category does not appear.
Do I have to turn off the disk cache on the Bot machine?
We recommend it on any machine where the Bot is left enabled. Empty the cache under Media & Disk Cache, then clear Enable Disk Cache if the machine is used mainly for automation. If you also design on that machine or need the disk cache, leave it on and clear it manually after each job the Bot runs.
Did After Effects 24.3 change how the purge preference works?
Yes. Before 24.3, the Templater preference Purge all memory caches after each job also cleared the disk cache. Since 24.3 it no longer empties the disk cache after each processed job, though it still clears RAM, image, and undo caches. That is why we advise handling the disk cache separately through the Media & Disk Cache preferences.
How often should the Bot check my Google Sheet?
The default is every 0.5 minutes, or thirty seconds. The minimum is 0.3 minutes for Google Sheets and 0.2 minutes for JSON sources. Our documentation warns that setting the value too low may cause After Effects to behave unexpectedly, so stay near the default unless you have a reason to change it.
How do I run a very large data source without running out of memory?
Mark the rows ready by setting the render-status value, then keep the setting Allow up to ___ rows to be queued at one time at its default and maximum of 20. The Bot processes the rows in batches of 20. Our documentation uses 10,000 rows as the example and says the cap reduces the memory Templater needs.
Will these changes make my Bot faster by a certain percentage?
We do not publish a percentage, and nobody can give you an honest one without your project. The changes keep a long-running Bot from slowing down because of cache buildup. To see what they do for you, record the time per row after several hours of running, apply the changes, and record it again.
