Making-Of | How I prepare images for print
The current situation
With the end of 2022 I have reached a milestone I was working on for almost 2 years. Finally I am able to offer a selection of my images to interested people as high quality prints on fine art paper. The turning point in my efforts was a workshop I was attending in March 2021. The course’s goal was to teach the participants how to build an online print shop within 10 days, enabling them to offer their prints worldwide. Since I quickly realized that such a tight schedule wouldn’t do justice to my high expectations in relation to the outcome quality of the prints, I spent months on this critically important aspect.
Theory vs. Practice
Theoretically the print preparation task appeared to be an easy one: I was told to sharpen the images with Topaz Sharpen AI before upscaling them with Topaz Gigapixel AI to the desired size, that’s it.
If you don‘t examine the files too precisely, you will get the impression of surprisingly detailed and sharp high resolution images, which are well suited for large scale printing, but since I’m rather bad in not looking closely, some unpleasant and unacceptable flaws quickly caught my attention.
Not only did the upscaling of the already sharpened image result in strong and ugly lines along high-contrast edges, but sometimes the underlying algorithms of Sharpen AI and Gigapixel AI made, despite their overall good quality, some almost ridiculous mistakes in certain areas. Because these issues were absolutely unacceptable for me, I did not rely on this almost automatable workflow, but changed the order of processing (now sharpening after upscaling) and decided to manually take care of all the little troublemakers.
I’ll show, based on the example of my image “Veil of Water", how much effort I had to put in this task and why a 10-day challenge therefore turned into months of tedious retouching work.
“Veil of Water” | 2019
Step 1 - Upscaling
First things first: Topaz Labs officially recommends to sharpen the images before upscaling them. After many tests I found out that the reversed order gave me better results with my pictures - at least for my personal aesthetic taste.
Because it appeared plausible to me to enlarge the image before sharpening it, I exported a 16-bit TIFF from Lightroom (without the vignette and the split toning effect) and opened the file in Gigapixel AI. The first thing you have to do there is to determine the desired image size. Since I wanted to offer my prints up to a size of 36 inch - which is almost one meter on the long side - in the highest possible print resolution of 300 dpi (dots per inch), I needed files with 10800 pixels on the long side.
The according calculation is rather simple: Target size in inch * Print resolution in dpi = pixel count, which in my case is 36 * 300 = 10800. By the way: If you would like to get such a high resolution image directly from your camera, you will need a model with at least 77 megapixels.
Once the target size is determined, you will have to choose one of the four provided upscaling algorithms, which all have their individual strengths and weaknesses and were designed for different kinds of source material. Besides the standard settings, you will find one for architectural pictures, one for painted/computer-generated images and two others for low resolution or heavily compressed files. During my tests I found out that one of these algorithms usually gives me the best overall result, while others may deliver a more convincing outcome in certain image areas. Please have a look at the following pictures for a better understanding of this problem:
The standard algorithm delivers the best overall result, but has problems with the structure of the cobble stones.
The result of the “Very Compressed”-algorithm is far inferior, but the cobble stone area is better defined.
Instead of compromising by choosing one of the variants, I am rendering out both versions (sometimes even three or four) and combine them via layer masks within Photoshop.
Here’s an overview of the layer structure so far:
In total I have used 3 different upscaling algorithms for the image “Veil of Water”.
The layer with the best overall quality is at the bottom (Standard Upscale), followed by the one with drastic details (Sharper Upscale) and above these sits the version with a rather gentle enlargement algorithm (Softer Upscale).
Some mistakes were retouched on the top layer (Grobkorrekturen) and I have already tried to get rid of some distracting elements.
This is the mask for the layer ”Sharper Upscale“. The black areas represent the standard algorithm, replaced by a more detailed version wherever I have painted with white.
After having found the best upscaling variant for each area of the image this way, I flatten the layers on a new one and start doing some first, rough corrections on it. Once finished, I save the result as a TIFF file to sharpen it in Topaz AI.
Step 2 - Sharpening
Just like in Gigapixel AI, you will get a list of algorithms for a variety of scenarios in Sharpen AI, and I found the same issue here: There is always one algorithm that gives you the best overall results, but comes with flaws in certain image areas.
There are almost always some very soft spots, as well as oversharpened parts, but by rendering out multiple versions with varying degrees of sharpening and combining them via layer masks in the manner I have described above, you can get results that show a balanced sharpening impression.
Left: The layer stack after the sharpeneing process | Right: Mask view of the layer “Soft Sharpening”. The white areas represent areas (mainly water) I found to be too sharp, so I had to blend in a version with a much gentler algorithm here.
Step 3 - Individual corrections
The whole workflow I have explained so far is the same for every picture, but now it´s time for individual corrections, which may differ vastly in the kind of treatment and the effort you have to put in. For “Veil of Water” I had to do the following:
Layer “Frequenztrennung” (Frequency Separation): Due to this technique, which is usually used for retouching portrait images, I was able to clone in fine moss and leave details in areas that - despite being treated with the most aggressive sharpening method - didn´t show enough definition. The great advantage of using frequency separation is that you don’t have to worry about finding similar tonalities for the clone source and the target, since the structures and colors of the image are separated in two layers, hence the name of this technique.
Layer “Rauschen Add” (Noise Add): It sounds counter-intuitive to add noise when usually everybody talks about keeping the noise as low as possible, right? The denoising parameters in Gigapixel AI and Sharpen AI may work well for most of the image, but in certain areas they can lead to a slightly artifical looking appearance. By adding some digital grain this problem is easily fixable.
Layer “Störelemente” (Distracting elements): Although I have already taken care of the most obvious flaws right after the upscaling process, I still stumble upon tiny distracting mistakes and elements. Be it a rotten leaf or a twig that’s placed in an unfavorable manner: Time to get rid of all those little things. Here I am a pixel-peeper, here dare it to be! I literally scan the image systemically with 800% magnification and correct all the micro disturbers I find.
Layer “Abdunkelung Büschel”: A patch of moss that attracted a bit too much attention due to its brightness was darkened with the help of a gradation curve layer.
Layer “Color Grading & Vignette”: This is a merged layer containing all the previous corrections. I also reintegrated the color grading and the vignette that were disabled before the initial export from Lightroom.
As mentioned before, the amount of effort I have to invest in this post-processing step varies heavily from file to file. I am currently working on the image “The white Temple”, which needs a precise selection of the sky. For certain reasons I couldn’t go the usual way and do the job with the help of luminosity masking, but had to manually paint in the mask with a 2-pixel sized brush instead. On an image with 10800 pixels on the long edge this was an extremely tedious process, which took me more than 50 hours.
Masking nightmare: The selection of the sky was extremely challenging due to the fine details in the trees and bushes.
Step 4 - Test prints
Initially I thought I could skip the expensive test printing process because I had calibrated the monitor and profiled my printer/paper combination. Shouldn’t I get a decent preview of the outcome by softproofing the image?
Well, kind of. It works quite well in relation to the hue and the saturation of colors, but estimating the outcome brightness is a different story. Although I have already dimmed my computer display a good amount, the printed images tend to appear almost always darker than their digital pendants. I guess there is probably no way to get a 100-percent accurate preview, since the difference between a light-emitting display and light-reflecting paper, whose appearance varies drastically in changing light conditions, is just too fundamental.
To cut a long story short: The proof of the pudding is in the eating! First of all I print the original edit plus three more versions with varying degrees of brightening on one DIN A4 paper, choose a favorite and take notes of any occurring issues. After the necessary adjustments I repeat the whole process and get then at the latest a result that matches my expectations.
Finally, taking the determined modifications into consideration, I print the image full scale on DIN A4, before gradually moving to the larger formats up to DIN A2. Whenever I find flaws in the increasing detailed prints, I correct those in the underlying master file.
Test prints with handwritten notes
That’s the way the mentioned DIN A4 test prints with 4 variations look like.
Why all this effort?
You may ask yourself why I am putting all this seemingly unnecessary effort into my prints and how reasonable it is to zoom in to 800% to edit the tiniest micro details. Who, besides me and maybe a few other interested photographers, will ever look at these large scale prints from such a close distance to be able to appreciate the amount of diligence I am treating my images with?
Well, the answer is quite easy: It’s just the consistent continuation of the standards I devoted myself to.
I always try to do my very best in every single step of the image creation, be it the planning, the capturing or the post-processing. If I would apply other criteria in the final step, I felt like the early wake ups, the sleepless nights, the countless hours of postprocessing and last but not least the lost family time were spent in vain. My images aren’t perfect, but I can confidently claim I gave my very best to come as close as I possibly could.
This aspect is crucial to me, especially in relation to the print shop. Every potential customer of my fine art prints has the right to expect not only beautiful images on high-quality papers, but also the greatest accuracy during the production process.
Register to our email service, and you will not only get notifications on updates on this site, but also on offers from our shop - exclusively for friends of bilderschmied.com.