My Cartoons Were Featured at SAP Sapphire, Here’s How I Made them.

A few weeks ago, the content team for Sapphire Orlando and Barcelona reached out and asked me to help create a cartoon for SAP CEO Christian Klein’s keynote presentation.

On the theme of “change” it needed to feature two panels: in the first, a leader asks the crowd “who wants change?”, and everybody has their hand up, and in the second, the leader asks “who wants to change?” and everybody has their hands down.

As you can see on this site, I’ve created a lot of cartoons over the years, mostly simple vector outlines using Adobe Illustrator on Mac or Inkpad on iPad. The problem was that I’m not a very good artist, and this process took more time than I had to spare, so I had a long backlog of cartoon ideas that I didn’t get around to creating.

And then ChatGPT came along, with the ability to create cartoon drawings with just a few clicks, and I’ve been able to use that to create cartoons faster, with better artistic quality.

I know that many people feel that using such AI image generators is unethical, since it’s leveraging the art of others, without any payment or acknowledgement. Personally, I’m conflicted. For millennia, art has thrived by drawing inspiration from, and remixing the work of, other artists. As in other areas, new technology has often come along that does what humans have always done, but now does it at scale. Society gains, but at the expense of the individual artisans who had previously done the work.

Various courts around the world are wrestling with the issue of rights and payments. For now they have generally ruled that AI generated art is a legally acceptable transformation of the underlying images (based on the rulings that, for example, allow Google to provide you with thumbnails and serve up similar images during searches). And they have also ruled that the artwork created is not copyrightable—unless there is a substantial human contribution to the finished artwork (and what exactly that means is being debated by the courts).

In the meantime, I feel like I’m mostly using AI as a tool to get me where I already want to go, faster—a sort of powerful extension of the drawing programs I already use (Adobe products do this directly, using their Firefly engine, based on images they say have been more ethically sourced—but it’s bad at cartoon art).

I come up with an idea, use ChatGPT to come up with possible images to illustrate it, choose among them, and then I use a combination of cutting and pasting and redrawing and adding text to create the final cartoon.

While we wait for the legal and moral boundaries to be drawn more clearly, transparency is important, so here’s an idea of the workflow that went into creating the cartoons for the keynote presentation.

First step: I asked ChatGPT to create a finished cartoon for me—explaining the two panels, giving the wording etc. But the current image engine that is used, DALL-E 3, is currently incapable of following the instructions (the new GPT-4o is multimodal, and may be able to do better when it’s available; currently it only does the text side of things).

Here are some of the results—as you can see, it struggled, mixing up the concepts. None of this was useable.

So instead I asked ChatGPT to create some images just of the leader and the crowd, with their hands up, and then asked it to try to create the same image with the their hands down. This created some OK images, but it proved impossible to get everybody’s hand in the air, and the second image was never close enough to the original to be useable.

I eventually realized I would have to create a simpler image that I could overdraw with what I needed. After generating dozens of different examples, I decided to use this one:

And based on that, with cutting and pasting and overdrawing, I created a two panel cartoon:

But for Christian Klein’s keynotes, they wanted to show the cartoons in a wider format, and one after the other, so after some new prompting using the original image I ended up with this image generated by ChatGPT:

I then cut and paste and repainted etc to create these two panels:

And for my own amusement I added a third and more cynical panel, which I feel may be a truer representation of the challenge in many (most?) settings:

It was then pointed out that this cartoon is far from being inclusive (i.e. it’s too much like the reality of most organizations!).

So I asked ChatGPT to make the speaker African-American, it came up with this:

I merged the two men and their stances to create the two panels:

And then I did something similar for a woman presenter. ChatGPT generated this:

I took this image, but merged it with the podium from the previous image, added a nose, redrew the hand in the second position, changed the smile, etc. (I would have added more women in the foreground of the crowd, too, but I was running out of time).

Christian’s team went with the first two, and they were shown in both Sapphire Orlando and Sapphire Barcelona:

[Bonus points if you notice something “interesting” about how the images were cropped in the final presentation]

Finally, to circle back to the ethical issue: as you can see, there was a lot of human intervention to create the final cartoons: selecting which images to use; mixing and repositioning the elements from different generated images; cutting and pasting elements (e.g. repeating the raised hands); manual extension and recoloring of the artwork.

But does this mean that the final images are now ethically acceptable, and/or copyrightable? I simply don’t know. One thing is clear: somebody else following the same method and using the same AI image generators would have ended up with a completely different cartoon, through their personal, subjective choices…

In any case, like my other cartoons, and to the extent I can legally actually let you do so, you can use them freely in your presentations if you would like (although please drop me a line to let me know what you’re doing: It’s always a pleasure to see what people are using them for…).