The Test That Started With a Deadline and 340 Product Images
A client sent over 340 e-commerce product shots last spring. Mixed lighting conditions, inconsistent backgrounds, varying levels of lens softness depending on which photographer was on set that day. The brief was clean white backgrounds, sharpened details, color-corrected to match the brand’s existing library. Standard stuff, except the turnaround was 48 hours.
I had two plugins already licensed and sitting in my Photoshop toolbar: Luminar Neo and Topaz Photo AI. I’d used both in lighter rotation for months, but I’d never actually put them side by side under real production pressure. That deadline forced the comparison I’d been putting off. What I found reshaped how I route work through my system.
What These Tools Are Actually Doing to Your Pixels
Before I get into the results, it’s worth understanding the difference in approach between these two tools, because it explains almost everything about where each one succeeds and fails.
Topaz Photo AI is fundamentally a sharpening and noise-reduction engine built on convolutional neural networks trained on millions of images. When you run Sharpen AI on a slightly soft product shot, it’s not applying an unsharp mask. It’s predicting what the in-focus version of that image probably looked like and reconstructing detail at a pixel level. The same logic applies to Denoise AI and the upscaling module. Each process is analyzing subject type (face, landscape, product) and motion blur characteristics separately.
Luminar Neo takes a broader editorial approach. Its AI tools handle sky replacement, background separation, portrait retouching, and relighting, but it’s not primarily a technical correction tool. It’s closer to a creative assistant. The underlying models are doing scene segmentation and layer compositing automatically, which is impressive, but it means the tool is making more assumptions about what you want.
That distinction matters enormously when you’re doing production work at volume versus personal creative projects.
Topaz Photo AI in a Commercial Context: The Settings That Actually Matter
For those 340 product shots, Topaz handled the sharpening and noise pass. Here’s how I configured it for a consistent batch result.
Inside the Topaz Photo AI interface, I ran the Auto Pilot feature first on a representative sample of about 20 images - roughly 6% of the batch. The model correctly identified “product” as the subject type in most cases, though it occasionally defaulted to “other” for items with complex textures like fabric swatches. For those, I manually set Subject to “other” and pulled Strength back to around 45 to avoid over-processing the texture into something artificial-looking.
Noise reduction was set to Normal, Strength at 40, and I left Detail Recovery at its default 50. For sharpening, I used the Standard model with Strength at 55 and Lens Blur Detection enabled. Processing time for the full batch, routed through Lightroom as the external editor and exported back as 16-bit TIFFs at 300 DPI, was approximately 4.2 hours on my current workstation (Ryzen 9 7950X, 64GB RAM). That’s a lot, but it ran overnight unattended. Total cost of the tool at time of writing is $199 as a one-time purchase, which is remarkably reasonable for what it does.
Where Luminar Neo Fits and Where It Doesn’t
I want to be fair to Luminar Neo because I think it gets unfairly dismissed by technically-oriented retouchers. For the right work, it’s genuinely fast and capable.
The background removal tool (called Background Removal AI inside the module) performed well on products with clean silhouettes - bottles, shoes, electronics. On items with fur, loose fabric edges, or transparent material, it needed manual cleanup, which I did using the Erase and Restore brushes at around 15-20 pixel radius. For a photographer doing direct-to-client work where the goal is a good result quickly, this workflow is entirely serviceable.
Where it fell apart for me was in the color accuracy of its relight tools. The AI Portrait Bokeh and Relight AI features shift color temperature slightly as a side effect of their depth-mapping process. On skin tones in lifestyle shots this might be unnoticeable, but on a product that has a specific brand color - say, a packaging shade that the client’s brand standards document specifies to the hex value - even a 150K color temperature shift becomes a rework conversation. I confirmed this by sampling identical areas before and after the relight pass using the Eyedropper tool. The shift was real and repeatable.
Luminar Neo’s subscription runs $9.95 per month or $79 per year. For hybrid shooters doing both personal and commercial work, that pricing structure makes sense. As a pure production tool, the color reliability issue is a dealbreaker for my specific clients.
The Spreadsheet Doesn’t Lie
I track time savings from every tool and action in my stack. This habit started practically - I wanted to justify software costs to myself in a concrete way - and it’s now a standing column in a spreadsheet I’ve maintained for years. The current total across all tools and actions is somewhere past 2,400 hours saved since I started keeping records.
After that spring batch job, I added a row for Topaz Photo AI: 11.3 hours saved on that single project versus my previous sharpening workflow using High Pass layers and Smart Sharpen at a frequency-separation level. Luminar Neo’s background removal saved me 2.8 hours compared to manual masking on the cleaner product shots. Those numbers are conservative. I’d rather undercount than convince myself something is working better than it is.
The Routing Logic I Use Now
For production work - e-commerce, advertising assets, anything with a client color brief - Topaz Photo AI is the right tool and Luminar Neo is not in that chain. For personal projects, editorial-style work, or when I’m handing off a rough composite to someone else for final polish, Luminar Neo’s creative speed is genuinely useful.
The mistake I see retouchers make is treating these as competing products when they’re actually solving different problems. Knowing which pipeline a job belongs in before you open the file is the decision that saves or costs you hours.
Comments (4)
Printing this out and pinning it next to my monitor. That good.
Been doing this wrong for years apparently. Thanks for the wake-up call.
This should be required reading for anyone starting out.
Love this. I referenced a similar technique in one of my recent posts. Always good to see other perspectives.
Leave a Comment