Disclaimer: All base images below were generated by AI. I am aware of the ethical considerations here. I am sharing these for entertainment purposes only; these are not commercially available. If you feel any of these images infringes on your or someone else’s copyright, please let me know and I will proceed accordingly, including deleting the image if appropriate.
It started with the Fool card, as it generally does.
I was playing with the Bing Image Creator (which uses the Dall-E AI engine) and decided to see how it would do with the Fool card. This was my prompt: “a harlequin is dancing near the end of a cliff. at his feet is a dog jumping playfully. it is a sunny day.” I also added a few styles: Watercolor, oil painting, art deco.
These are the images from that prompt:
Over the next two days (May 5 to May 7), I made the rest of the Major Arcana like this. Some of the images were awesome; some of them were awful. The Wheel of Fortune was particularly challenging (prompt: “a carousel with various animal seats, including a lion, an ox, an eagle, a snake, and a devil. people are riding the seats.”): AI got the general concept of the carousel just fine, but the animals were generally overcrowded, headed in both directions, and wildly mutated.
I then set the images aside, not sure what to do with them.
Months later, I decided to make the minor arcana cards. I took one day per suit (November 28 to December 1, during Thanksgiving Break), using the image descriptions from Biddy Tarot as a basic inspiration and riffing where I felt inspired.
In a few cases, I had to deviate from the traditional card because of Image Creator’s content restrictions. This was especially true of the Eight and Ten of Swords. Image Creator was willing to blindfold a woman, but not tie her up (understandably). My compromise prompt: “a blindfolded woman in a long gray dress is standing in the middle of a field, surrounded by upright swords.”
Also, I learned that after about three, it was pointless to include a number; Dall-E can’t count.
Now I had 78 sets of images and an urge to print them out, laminate them, and make my own cards. But encouraged by my spouse, I wondered how much it would cost to have a publisher make them correctly.
It turns out: Not that much. There are several companies that make custom card decks, and will even do one-offs at a decent price. I settled on Make Playing Cards, which has this sliding scale:
Because the images were AI generated and, hence, I wouldn’t feel comfortable selling these, I figured $23 for a single deck would be a really nice gift for myself. There is a significant minimum shipping cost of $12, but even after that, taxes, and a plain white box, the entire order came in under $40, still a fine price for a quality one-of-a-kind deck.
I set up a template in Canva. I do have an educator account through my school, but I wanted to make sure I had access after I leave that school, plus I’m not sure about the optics of making a tarot deck attached to a school account (even if I set that up myself and it’s free), so I created a free Canva account on my personal email.
Even with the limitations of the free version, though, I didn’t have a problem creating my cards. I started with a template card:
The template is 897px by 1497px at 300dpi, per MakePlayingCards’s default specifications. I also made sure there was a 36px border where there wouldn’t be any text or images in the “bleed area”, which would be cut off.
For the main image, I used these specs:
I cropped the image to these parameters. The border is a solid 10px line with a corner rounding of 15. I figured this would look nicest with the corner rounding on the cards themselves.
The text on the bottom is ITC Benguiat 60pt, y=1326px, horizontally centered, with the splice effect:
The set up easily took longest on the first card. Once I was happy with the layout, the rest of the deck was a matter of:
- Duplicate the existing card
- Choose my favorite image
- Crop and position
- Perfect using the settings
- Change the text
I also color-coded the suits, so there was a little work there, too. I also added a keyword to each one:
For the back, I chose an image I’d created in AI during a separate session:
which cropped to this in Canva:
which was further cropped by MakePlayingCards to:
Make sure to especially compare the cropping of the Canva image to the MakePlayingCards image.
I made sure to name each page in Canva so that those names would carry through to the file names. When I was done with the 79 images (having done a few quality checks along the way), I used Share > Download > JPG. Because I was using the free version, I couldn’t change any of the settings, but that was fine.
I uploaded the images to MakePlayingCards, following the prompts. That’s a drag-and-drop process that only took a few minutes; the bulk of the work was making sure the images would come through correctly, and not cut off any important stuff.
I placed my order on December 9, it was shipped on December 15, and I received it on December 23. Fine turnaround for a print-on-demand deck of cards. Final price: $37.38 (including standard shipping and taxes, and a plain white tuck box).
I was nervous about the quality, and I didn’t need to worry. I got a deck of professional-quality cards, standard thickness with a subtle level of gloss. The backs are absolutely identically centered and cut, so it’s not possible to tell the cards apart. Here is the back and the Fool card, scanned in:
The print version is slightly darker than the JPG images, but that sort of variation is not surprising when going from digital to print.
Based on this experience, I can definitely recommend both MakePlayingCards and the idea of making a custom deck (Tarot or otherwise).
Here is the final deck.
