Chelsea Jones creates AI art in the form of AI-enhanced oil paintings with female characters. But how does she like to use AI, and how will her AI-generated art look in print? We made a full-page image of her Woman sitting on a couch, in offset print on uncoated magazine paper.
I’m a product designer and artist, currently living in Atlanta, Georgia. I started working with AI for a number of reasons but at the forefront was the sheer intrigue and fascination with the potential this technology offered. AI has provided me with an expansive canvas for creativity. It has allowed me to delve into varied mediums, experiment with diverse styles and grapple with big ideas – all without having to worry about wasting money and resources. This is groundbreaking in itself.
My process of creating AI art involves a combination of different tools, each with a distinct purpose. I start off with ChatGPT to help me solidify abstract concepts. For the initial creation process, I rely on Midjourney which has a certain artistic flair that resonates deeply with my style. Once the initial artwork is generated, I move to Photoshop for the final touches. I’m particularly drawn to creating AI enhanced oil paintings that feature characters who resemble me and share similar stories. I utilise a blend of approaches; sometimes I start from scratch, and at other times, I leverage existing images. For structuring the subject and the scene, I employ my own tested prompts to maintain control over the composition’s overall layout and narrative trajectory.
AI technology has been an absolute game changer in enhancing my creativity, acting as a catalyst that accelerates my artistic process. I’m now able to rapidly iterate through ideas, experiment with different mediums, and bring to life the myriad concepts that cross my mind. Moreover, the freed-up time has allowed me to delve deeper into big ideas. AI acts as my collaborative partner, assisting me in translating these abstract concepts into tangible, evocative pieces of art.
Chelsea Jones, @kinleycollective.co
How were the warm colours and skin tones handled in prepress – did the image contain enough information for a great reproduction? We asked pre-press technician Jens Jörgensen, OTW, to share his best tricks from his job behind the scenes.
It's not uncommon to experience a "flatter image" when printing a colour-rich AI image, and it's important to understand why, to get the best result in print. Read more about colour conversion from RGB to CMYK here, and about using the correct ICC profiles here.
When an image contains large areas of a single colour, for example red, green or blue, the colour nuance may become a bit too much simplified when converted to CMYK, cyan, magenta, yellow and black. The four CMYK colours are used for the four printing plates in an offset printing press, and it depends on the exact nuance how many of these basic colour plates will be needed to get the final colour.
It's never ideal to use full colour on all plates at all times – this would result in subdued dark brownish colours and a waste of ink. But in some cases it might be a good idea to add a small touch of a colour from another printing plate to make a colour really pop. In this image, the orange-red pillow and the footstool were both slightly enhanced for print by adding some complimentary cyan after the colour conversion, to match with their colour-rich background.
Contrary to some preconceptions on the market, both light and dark skintones reproduce very well on uncoated papers. The raw, uncoated surface of the paper brings a natural feel to the image, where the skin looks smooth and soft.
It's important not to make the image too dark. An image on an uncoated matt paper may consume more ink than an image on a coated paper or glossy paper. There is also a difference in light reflection, because when an image is printed on a paper with high gloss, more light is reflected back from the surface. Therefore, always take an extra look at hair and skintones and check whether there is a need to make these areas in the image slightly lighter, so that contrast can be upheld. Often, using the correct ICC profile is enough.
In this image, a few light corrections were made for the face, hands and hair of the AI-generated model, and also in the areas directly behind her, where the background varies in dark hues of green.
I'm very happy with the print result, the rich colours and the natural reflections of light in the AI model's face as well as on the pillow on the couch. Also the reflections in her hair come out well; they look even more natural in print than on screen.
When we make magazines ready for printing, we always order a colour proof or a contract proof from the printer and examine this as a step on the way to the final result. In this case we can all see the finished result in the printed magazine, which can be ordered from Holmen here.
Writing "*/Imagine:" followed by a prompt initiates the image generation in the program used, and some of the AI prompts for making this image were:
The image was delivered in a JPEG format, packed in a size of only 4 MB, which seemed small at first, but it opened up in 48 MB to a size of 1.4 x 1.4 m in a 72 dpi resolution. This was definitely large enough to offset print in full magazine page size 200 x 265 mm with 300 dpi.
This image looked small for a full-page print when it was delivered from the artist, but don't judge from the size of an unopened file. What it really comes down to is whether the motive contains enough information to be reproduced well. You must open the image in an image program and examine it – does it look good? Then check the resolution: set the planned size for printing in inches and see whether the image output will contain 300 pixels per inch. Read more about sizes and formats for digital images here.
The warm orange-red colours contrasting with the green make this image special. They can use a bit of treatment to really pop in print. The prepress technician also made an extra look at the the hair and skintones to make sure all details come out well.