The job came in on a Thursday. Five hundred product shots, all needing background removal, a color grade, a sharpening pass, and export at three different sizes for web, print, and social. The client wanted them Monday morning. A few years into running my own consultancy, I’d learned to stop quoting turnaround times before I looked at my action library. I checked, confirmed I had what I needed, and told them Friday afternoon. I delivered Friday at 2pm, went to my kid’s soccer game, and billed for eight hours instead of the forty it would have taken me in 2012.
That’s not a flex. That’s just what happens when you treat batch automation as infrastructure rather than a trick you use occasionally.
Why Photoshop’s Batch Engine Is More Powerful Than Most People Use It
Photoshop’s batch processing, found under File > Automate > Batch, is not new and it is not glamorous. What it is, though, is a full execution engine for any action in your library, run against any folder of files, with granular control over how errors are handled, where files are saved, and what naming conventions get applied on output.
Most people use it like a vending machine: pick an action, point it at a folder, hit play. That works. But the real power is in the override checkboxes and the destination settings. When you check “Override Action ‘Open’ Commands,” Photoshop ignores any hardcoded file opens inside your action and replaces them with each file from your source folder, one at a time. Same logic applies to Save commands. This means your actions don’t need to know anything about your folder structure. They just need to do the work. The batch engine handles the logistics.
Combined with droplets (File > Automate > Create Droplet), you can package an entire workflow as a drag-and-drop application sitting on your desktop. Drag a folder onto it, walk away. This is what I had running on a Saturday afternoon when I built the system I eventually used on those 500 product shots.
Building an Action That Can Actually Scale
A batch-safe action is different from an action you record once and use on a single file. The main failure points are: hardcoded canvas sizes, pixel-based transforms that don’t scale across different source dimensions, and color profile assumptions.
For product photography, I run three separate actions chained through a droplet rather than trying to build one monolithic action that does everything. The first action handles background work and color correction. The second applies sharpening using a High Pass layer set to Overlay at around 60-70% opacity, with the radius adjusted to 2.5 pixels for web output. The third action handles resizing and export: 2000px on the long edge for print (at 300 DPI), 1200px for general web use, and 1080px square for social, all exported as JPEGs at quality 85, which keeps files under 400KB without visible compression artifacts.
The key to making these scale is using percentage-based transforms and conditional steps where possible. When the source files are inconsistent, which they always are from a live shoot, I include a flatten and canvas-normalize step at the top of my first action. Everything gets leveled to the same starting point before any real work happens.
The Spreadsheet I Have Been Keeping Since 2017
I track every significant action I build in a spreadsheet. Not because I am compulsive about data (I mean, I am, but that is not the point) - but because knowing the ROI on a tool changes how you invest in building it.
The batch system I built for product photography has logged over 340 uses since 2017. At an average of about 80 files per batch run, that is roughly 27,000 individual images processed automatically. I estimate manual processing at 4 minutes per image for the full workflow I described above. That is 1,800 hours of work that happened without me. The spreadsheet currently sits at 2,400+ total hours saved across all my actions combined.
I bring this up not to show off a number but because it reframed how I think about spending time on optimization. If I spend a Saturday afternoon, call it 6 hours, building and testing a batch workflow, and that workflow saves me 4 minutes per image, I break even after 90 images. For a consultancy doing regular e-commerce work, 90 images is a slow week. The math is almost always in favor of building the system.
Where Batch Automation Actually Breaks Down
Batch is not magic. It fails predictably in a few specific scenarios, and knowing them in advance saves you from discovering them mid-job at 11pm.
Layers with dynamic content, particularly smart objects linked to external files, do not play well in batch. If your action references a linked asset that is not in the expected location, the batch will error out or silently produce a wrong result. I flatten and embed everything before any batch job runs. No exceptions.
Conditional actions (via Image > Variables or scripted conditionals) also behave inconsistently across file types unless you have tested across your actual source format. I do all my testing on a sample folder of 10 files pulled from real client shoots before I commit to a full run. That 30-minute test saved me from a catastrophic export mistake earlier this year when I caught that my CMYK color profile setting was overriding a client’s specific Fogra39 requirement.
And if you are working with raw files, process them in Lightroom or Camera Raw first with sync settings applied across the selection, then export as 16-bit TIFFs into your batch folder. Trying to batch raw files directly through Photoshop introduces one more layer of inconsistency you do not need.
The One Thing That Separates Automation That Holds Up From Automation That Breaks
Build your actions to be ignorant of anything they do not control. File locations, naming, color profiles, canvas dimensions, those belong to the batch engine and your folder structure, not inside the action itself. The more context-specific assumptions you bake into an action, the more it becomes a single-use tool instead of infrastructure.
Every hour you spend making an action genuinely reusable pays you back faster than you think.
Comments (3)
Finally someone explains this in a way that actually makes sense.
Been doing this wrong for years apparently. Thanks for the wake-up call.
Never thought of approaching it this way. Really creative.
Leave a Comment