When I started this blog many, many years ago now I had a simple goal. My goal was to show consumers what was available in the coloring world so they could make an informed decision about the value of a product before they purchase it. This value proposition continues to be my goal and focus with this blog. With that in mind, it’s time to talk about the new player in the coloring world – AI-Generated Art.
AI Generated Art
What is Art?
What Art is Inside Our Coloring Books?
Before we dive into the world of AI Generative art let’s talk about the art that is already in our coloring books.
There are 4 main types of art that you will find in your coloring books:
- Hand drawn – this art has been hand drawn either on paper or a tablet. The artist guides the tool to create the art.
- Photographic – a photograph has the color desaturated to form a greyscale image. This differs from artists that create their greyscale images by manually adding shading.
- Public Domain – images available to the general public not protected by intellectual property law. This could be because the copyright in the images has expired or for many other reasons. An example is that by Sir John Tennniel who illustrated Alice in Wonderland. Several coloring books include these illustrations
- Computer Generated – software can create images like clip art, mandalas, kaleidoscopes, patterns, spirographs, and more that can be used in coloring books
There is now a new player in the coloring book world – art created by artificial intelligence (AI art).
Overview of AI Art and Hand-Drawn Art
What is Handdrawn art?
Pros of Hand drawn art
- Style: Hand-drawn designs offer a completely unique quality because the artist has control over the tool used to create the art. This means that with each stroke and line, their technique and style is being transferred. This ability allows them to completely manage the process, which is unique to them, and emphasize particular elements, line weights, create texture and shadows, perspective, and composition. The combination of these skills and experience allows the artist to develop a style that is unique to them.
- Personality: Illustrations created by hand often offer insight into the artist’s personality and life. This connection is often felt by consumers and can transfer an emotional response and a more meaningful coloring experience
- Unique: As well as their own style hand-drawn art is created on a physical medium rather than a tablet is one of a kind. This makes it unique and collectible as an original piece of art.
Cons of Hand drawn art
- Time: Hand-drawing designs can be an incredibly time-consuming process. It requires more physical effort and precision in the creative process as well as in correcting any errors. This makes it difficult for artists to produce a large number of designs in a short amount of time. It also makes it difficult for an artist to market their brand, and products due to the time spent in the creative stage.
- Editing: There are limited options to revise the art once it has been created without the difficulting of erasing parts of an image
- Design Block: Artists may often visualize the way they want an illustration to look but may not be able to find suitable reference pictures for inspiration or may spend time trying to execute the design but it does not meet their expectations
- Ability: In order to hand draw art considerable skill, practice, and experience is required which can take years and even decades to perfect
What is AI Art?
What is AI
AI-generated art refers to art that is created by using artificial intelligence techniques such as machine learning, computer vision, and natural language processing. In this process, a machine learning algorithm is trained on a large dataset of images or other forms of artistic content to learn patterns, styles, and features. Once the algorithm is trained, it can generate new art based on its learned understanding.
AI-generated art can take various forms, including images, videos, music, and even text. Some examples of AI-generated art include computer-generated paintings, music composed by AI algorithms, and videos that are created using generative adversarial networks (GANs).
In this article, we will only be discussing the use of AI art as related to images.
Google launched DeepDream in 2015, and hit a new peak last year when Stability AI rolled out its latest release of Stable Diffusion, the language model behind AI Generators like NightCafe and Lexica. Stable Diffusion quickly joined the ranks among Midjourney, DALL-E-2, and Niji Journey as an open-source language model for developers to create and experiment with AI art tools. Along with the release of OpenAI’s ChatGPT, AI is all anyone in the tech world can talk about.
As a result, AI art generators like LENSA and Artbreeder got into the game, enabling users to easily splice and edit their artwork, generate avatars, and collaborate on art projects. Software like Photoshop and Figma began integrating AI into their features. Novice designers can now create professional-grade designs in minutes with AI-powered freelancer tools like Canva and Adobe’s Sensei.
You can hear Fei-Fei Li explain this process in the Ted talk video below
This video simply explains how the image generators that we know today were created.
You can read more of the history of AI art in a timeline format here.
Pros of AI Generative Art
- Faster: Generating designs using AI algorithms is a much faster process than hand-drawing each design. This allows publishers to produce more adult coloring books in less time, providing a wider selection for consumers.
- Creativity: By using AI algorithms, artists can generate designs that they may not have thought of on their own, providing a source of inspiration, new ideas, and diverse designs
- Ability: Creators do not have to study art and practice art techniques, skills, and composition learning to create art
Cons of AI Generative Art
- Personality: As the designs are created by AI algorithms there is often a lack of the personal touch between the creator and artwork which can make the imagery less meaningful to some consumers.
- Limited Control: Although the algorithms generate designs based on the prompts of the creator there is little control over the output. This means that the result may not correspond to their vision and may need to be reiterated over and over in order to achieve the desired result
- Technology: AI generative art relies on technology, making it impossible to create designs without access to technology.
- Style: Can be difficult due to the lack of control, to develop a personal style that would make the creator and artwork more collectible and brandable
- Intellectual Property: A potential drawback of AI-generated art is the challenge of obtaining intellectual property rights for the generated output. This could result in your images being used by others without any legal recourse, especially if they excel in marketing and promotion.
Is AI-Generated Art Really Art?
Opinions are divided on this question.
Some believe that AI-Generated Art is art because:
- it improves based on its learning over time like humans;
- output is based on human emotions, desires, and imagination; and
- the technology used is merely an evolution of art techniques and tools
Many people consider that the use of technology without the ability, skill, and experience of an artist cannot produce images that should be classified as “art”.
There are others though in the art world that have embraced AI Art. Many see it as another tool that can be used by artists and like humans it is inspired by other works it has seen to produce its own original pieces.
AI Art Winners
Last year Jason Allen submitted this picture, “Théâtre D’opéra Spatial,” to the Colorado State Fair in the “digital art/digitally manipulated photography” division.
He had created the picture using the AI Art Generator, Midjourney.
The picture ultimately won first prize. Mr. Allen took home the blue ribbon and $300 for his win.
It is believed to be one of the first A.I.-generated images to win a prize.
The win was controversial with many believing that the creation of the picture and his ultimate win was “cheating“.
It is easy to understand why artists led the backlash. In a short period of time, in comparison, to a digital artist, Mr. Allen was able to create an award-winning picture. A traditional artist would have spent months creating the image and years honing their skills in order to do so.
He said of his win: “I couldn’t believe it,” he said. “I felt like: this is exactly what I set out to accomplish.” and added, “This isn’t going to stop,” Mr. Allen said. “Art is dead, dude. It’s over. A.I. won. Humans lost.”
More recently. Boris Eldagsen a German photographer submitted a black and white called “The Electrician” to the Sony World Photography Awards and won. He declined to accept the award stating that he had only entered “as a cheeky monkey, to find out, if the competitions are prepared for AI images to enter. They are not,”
“We, the photo world, need an open discussion. A discussion about what we want to consider photography and what not.” World Photography Organisation (WPO) said that Eldagsen told them he had “co-created” the image using AI before he was announced as the winner.
It seems that there was some misunderstanding about the AI aspect of the entry with WPO saying “Following our correspondence with Boris and the warranties he provided, we felt that his entry fulfilled the criteria for this category, and we were supportive of his participation.” going on to say “Given his actions and subsequent statement noting his deliberate attempts at misleading us and therefore invalidating the warranties he provided, we no longer feel we are able to engage in a meaningful and constructive dialogue with him”.
AI Art Galleries
As well as winning prizes at local fairs and shows, AI Art has also been exhibited in Art Galleries.
In 2022 Bitforms Gallery in San Francisco held an exhibit called “Artificial Imagination” based on artwork from well-known AI art generator, Dall E. The gallery owner is known for combining art with technology and considers AI to be just another tool in the artist’s toolbox.
Recently in Amsterdam, The Netherlands the first AI Art Gallery opened called the Dead End Gallery. The artwork is created by fictional “AI Artists” and features work from artists that are certainly not perfect, with 6 figured hands featured. Notwithstanding that, AI artist Irisa Nova the Gallery said the, “estimated asking price for the curator could vary between €3,000 and €10,000” for each piece of art.
The members of the AI Art Gallery were inspired by Obvious Art to delve into the work of AI. Initially, it started as a project when they were at Cambridge University but within months the three childhood friends had a business. The Gallery aims to provide affordable art to people. Their works are printed on canvas and come with a digital file and a certificate of authenticity. You can read more at AI Gallery.
As well as galleries you can also buy and sell AI artwork at dedicated marketplaces like Artsi.ai
AI Art is mainstream now and no longer novel. The pace at which it has been adopted by the general public is astounding and like it or not, it seems that it has been accepted as art.
How do AI Generators Work?
Training the AI Generator
There are a number of AI Generators so this information is general rather than specific to a particular AI generator.
AI models, are typically trained using a process called “machine learning”. Here’s a simple explanation of how this process works:
- A large dataset of images is collected. This dataset might include thousands or even millions of images, depending on the complexity of the AI model.
- Next, the AI art generator is initialized with some basic parameters and architecture. These parameters and architecture will determine how the model processes and generates new images.
- The AI art generator is then trained on the dataset of images. During this training process, the model is shown a series of images from the dataset and asked to generate new images that are similar in style or content. As it generates these images, the model is given feedback on how well it’s doing, and it adjusts its parameters and architecture accordingly.
- This process of training and feedback continues for many iterations until the model is able to generate images that are similar in style or content to the images in the dataset.
- Once the AI art generator is fully trained, it can be used to generate new images on its own, without any input from humans. These new images might be completely original, or they might be based on some input or prompt provided by a user.
The actual process of training an AI art generator can be much more complex than this, and there are many different techniques and algorithms that can be used to train these models
Lets’s dig a bit deeper and examine the data set of images
Step 1 – Data Set Training
A large dataset of images is collected. This image data can include paintings, photographs, illustrations, sculptures, or any other type of art.
The images in the dataset might be of many different styles and subjects, depending on the goals of the AI art generator. For example, if the goal is to generate realistic portraits, the dataset might include thousands of photographs of people’s faces. If the goal is to generate abstract art, the dataset might include many different types of colorful and patterned images.
The datasets may be curated by humans who select and organize the images based on certain criteria, such as artistic style or subject matter.
Some popular datasets that have been used to train AI art generators include:-
- the ImageNet dataset from 2010. This data set contains over 14 million images of objects in many different categories. The research in this data set provided by Princeton and Stanford Universities is available for free use for non-commercial purposes;
- COCO dataset, which has over 330,000 images of everyday objects and scenes. COCO is sponsored by some major players including Microsoft and Facebook; and
Laion 5B has over 5 billion images in its dataset. Laion 5B is a not-for-profit research group attached to a University in Germany. Its sponsors include Hugging Face, Doodledot, Stability, and the eye which are major players in the AI community.
Collecting and organizing the data set ensures the AI can produce diverse artwork. By learning from a large and varied dataset, the AI art generator is able to create new images that are similar in style and content to the images in the dataset, while also introducing new and different variations.
Now here’s the first problem
The images used for these data sets were not donated, gifted, licensed, or released to the AI generators by the copyright owners. They were scraped from the internet as training data for the AI to learn from. Theoretically, the picture that you took of your cat could be in this scraped load of images as well 🙂 For example, when you upload a picture of a cat on the internet and call it a “cat” the AI learns “this image is what a cat looks like” and so on. It can then learn to identify the images through this caption or text.
Artists who rely on the sale of their artworks found their works included in these data sets. At the time they were not presented with an option to be included or excluded or a way in which to remove their artwork from the training in research that they did not agree to be part of.
Step 2 – Rules & Parameters
The “parameters” of an AI model are the variables that the model uses to make predictions or generate new data. In the case of an AI art generator, the parameters might include things like the size and resolution of the generated images, the colors and shapes used in the images, and the level of detail in the images. These parameters can be adjusted to create different styles of art.
The “architecture” of an AI model refers to the overall structure of the model and how it processes data. There are many different architectures that can be used for AI art generators, but one popular type is called a “generative adversarial network” or GAN. A GAN consists of two neural networks that work together: a “generator” network that generates new images, and a “discriminator” network that evaluates whether the images generated by the generator are realistic or not. The two networks are trained together in a process called “adversarial training”, where the generator tries to create more realistic images to fool the discriminator, and the discriminator tries to correctly identify whether an image is real or generated.
Other architectures that can be used for AI art generators include “autoencoders” and “variational autoencoders”. These architectures also involve neural networks that are trained to generate new images based on patterns in a dataset, but they use different techniques than GANs.
Steps 3- 5 The Training Process
These images are usually used to train the deep neural network of the AI art generator by feeding them into the network, allowing it to learn patterns and features from them. During training, the images are typically divided into batches, and the neural network is updated in a process called backpropagation. This means that the weights and biases of the network are adjusted based on the error between the predicted output and the actual output. The process is repeated for many iterations until the network reaches a desired level of accuracy. After the training process is complete, the weights and biases of the neural network are saved as a model file, which can be used to generate new art based on input from the user.
You can really dive into how this process works by reading this article which visually explains how Stable Diffusion works
When an AI art generator has been trained, the images used for training are often not needed and are discarded or archived.
Ongoing User Training
How do you create AI Art?
Based on my experience, I will give you a brief rundown of how they work.
- Once you have set up your account follow the “newbie” instructions on the site for creating an image.
- Then all you need to do is think of an idea for an image that you would like to create. Now think of how you would describe that idea in text format.
- The text is how you instruct the AI generator to create your image. This is called a “prompt”. Translating your idea into a prompt called “prompt crafting” or “prompt engineering”.
The last picture looked the most like Milly so I regenerated that image. This is as easy as clicking a button.
Is this legal?
Perfect 10 v Google Inc
Perfect 10’s Argument
Google contended that it had a right to use the images on the basis of fair use.
The fair use defense permits the use of copyrighted works without the copyright owner’s consent under certain situations, which are decided on a case-by-case basis. The defense encourages and allows the development of new ideas that build on earlier ones and counterbalances the copyright law’s goal of protecting creators’ work products.
Legislation codified this law into the US Copyright Act at s107 which provides for 4 elements to be considered when deciding whether the defense of fair use applies.
The elements are –
1) the Purpose and character of the use, including whether the use is of a commercial nature or is for nonprofit educational purposes
2) the Nature of the copyrighted work
3) the Amount and substantiality of the portion used in relation to the copyrighted work as a whole
4) Effect of the use upon the potential market for or value of the copyrighted work
In summary, these factors provide copyrighted work can be used in certain circumstances like “criticism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, or research, is not an infringement of copyright”.
This case went to the Court of Appeal which ultimately decided that the way in which Google used the images for the purpose of search engine results transformed the images into a different purpose that “serves a different function from the original work”. Weighing up the benefit that Google provides to the public by providing image search they concluded that the use of Perfect 10’s thumbnails was fair.
In another similar case, again with Google
Authors Guild Inc v Google Inc
Google decided to digitalize and scan books so that the books could be searched on their google books search function. The whole book was not available for viewing just small snippets that included the search term or phrase that was inputted.
Author’s Guild Argument
The Authors Guild acting on behalf of authors sued Google for copyright infringement as consent was not obtained from the copyright holders of the tens of millions of books that were scanned by Google.
Just like in the previous case, the Court concluded that:
(1) Google’s unauthorized digitizing of copyright-protected works, creation of search functionality, and display of snippets from those works are non-infringing fair uses. The purpose of the copying is highly transformative, the public display of text is limited, and the revelations do not provide a significant market substitute for the protected aspects of the originals. Google’s commercial nature and profit motivation do not justify the denial of fair use.
(2) Google’s provision of digitized copies to the libraries that supplied the books, on the understanding that the libraries will use the copies in a manner consistent with the copyright law, also does not constitute infringement. Nor, on this record, is Google a contributory infringer. Accordingly, the court affirmed the judgment.
Does this apply to AI Generators?
Both of the cases above referred to scraping something that was already on the internet like AI Generators have done. Unlike the previous cases though, there are some differences with AI Generators. First, many are accessing a data bank created by a not-for-profit third party to create art, effectively putting a middleman between the user, the generator and the scraper. Secondly, the data that was scraped is not being used in its original format but is being used to generate something new and not a copy of an existing work.
As this is new technology and a very complex area. It is difficult to see how Courts will consider the legality of the usage.
Luckily we don’t have to wait too long.
The lawsuits have started
Until the legislation is changed or there is a precedent from the common law, the question of whether the scraping of images constitutes fair use or not is unknown.
Fortunately, there are many cases in the works that may be answering this question and others. Even, if they don’t, they will be providing some steps towards clarifying the issues.
Andersen v. Stability AI Ltd, U.S. District Court for the Northern District of California, No. 3:23-cv-00201
There is a major case at the moment in the US where three artists, —”Sarah Andersen, Kelly McKernan, and Karla Ortiz— filed a class-action lawsuit against Stability AI, DeviantArt, and Midjourney for their use of Stable Diffusion” a 21st-century collage tool that remixes the copyrighted works of millions of artists whose work was used as training data.
Anderson & Others Argument
The artists are claiming:
“direct and vicarious copyright infringement under 17 U.S.C. § 501; violation of the Digital Millennium Copyright Act, 17 U.S.C. §§ 1201–1205 (the “DMCA”); violation of Plaintiffs’ statutory and common law rights of publicity, Cal. Civ. Code section 3344; violation of Unfair Competition law, Cal. Bus. & Prof. Code §§ 17200, et seq.; and declaratory relief“.
Reading through the complaint, the main arguments are copyright infringement and a claim for derivative works. As set out in the complaint “Stable Diffusion uses the Training Images to produce seemingly new images through a mathematical software process. These “new” images are based entirely on the Training Images and are derivative works of the particular images Stable Diffusion draws from when assembling a given output. Ultimately, it is merely a complex collage tool. “
“These resulting derived images compete in the marketplace with the original images. Until now, when a purchaser seeks a new image “in the style” of a given artist, they must pay a commission or license an original image from that artist. Now, those purchasers can use the artist’s works contained in Stable Diffusion along with the artist’s name to generate new works in the artist’s style without compensating the artist at all”
Stability & Others Defence
It’s early days for the litigation at the moment and the defendants have yet to file their Defence, at the time of writing. It seems likely based on the cases previously mentioned that they will argue a fair use defense in respect of the copyright infringement.
A derivative work as defined in the US Copyright Act is ” work based upon one or more pre-existing works, such as a translation, musical arrangement, dramatization, fictionalization, motion picture version, sound recording, art reproduction, abridgment, condensation, or any other form in which a work may be recast, transformed, or adapted. A work consisting of editorial revisions, annotations, elaborations, or other modifications, which, as a whole, represent an original work of authorship, is a “derivative work”.”
Copyright owners are in control of allowing others to create derivative works. However, Stability and others would certainly have arguments that they could raise. Without knowing the inner workings and the technical side of the business, I imagine that they would argue that it is the original work that is being used but rather an algorithm and in any case such a small amount as to be irrelevant and unidentifiable.
The artists have also referenced their “style” in their complaints. Unfortunately, “style” is not protected by copyright law. If it was a direct copy of an existing image rather than using the artist’s style for a different picture then that is a different matter. That would be a direct copyright infringement. It does not appear that this is the case, from what we know so far, in this case.
At this stage, it is unclear how a court will interpret these issues as it is the early stages of the litigation.
I am not an AI expert. My limited understanding of how AI image generation works do not convince me the argument by the artists is accurate. Splitting hairs, but my understanding is that the images train the AI. This then marries the image to the text. Images are then deleted from the database and the AI works off the algorithm to make a creation and what it has learned usually through a “diffusion” process. This concept is different from what the artists are currently arguing. I am happy to be wrong, the tech subject is above my pay grade.
Without being privy to the ins and outs of this case, even if I am wrong on the techie aspect, I am of the belief that Stability and others will be successful if they use a fair use defense. My opinion is based on the previous cases quoted and the use of a third-party, not-for-profit “research” site to facilitate the data bank.
I’ll be eagerly keeping an eye out on these proceedings as no matter who is successful it will bring more clarity to this constantly evolving world.
Getty Images (US) Inc v. Stability AI Inc, U.S. District Court for the District of Delaware, No. 1:23-cv-00135
Earlier in February 2023 Getty Images a large stock photography site commenced proceedings against Stability.
Getty claims that Stability has infringed on its intellectual property rights in a number of ways. Stability has scraped images from Getty that are high quality, have valuable metadata, and are often watermarked. The high-quality images are particularly valuable to the AI training model because of this. Getty also claims that some images generated by Stability show the Getty watermark which could cause customer confusion.
Getty is asking for an injunction and claiming damages including future profits.
It is early stages. At the time of writing Stability has not filed a defense. The fact that Getty is asking for future profits in its damage claim makes me believe that Getty would be happy to license the images to Stability and work with Stability. I’ll update this section when we know further information.
Getty Images (UK) Inc v Stability AI (UK, High Court, London)
On 17 January 2023, Getty Images in the UK commenced proceedings against Stability AI claiming “claiming Stability AI infringed intellectual property rights including copyright in content owned or represented by Getty Images. It is Getty Images’ position that Stability AI unlawfully copied and processed millions of images protected by copyright and the associated metadata owned or represented by Getty Images absent a license to benefit Stability AI’s commercial interests and to the detriment of the content creators”.
In a statement, Getty Images stated that it “believes artificial intelligence has the potential to stimulate creative endeavors. Accordingly, Getty Images provided licenses to leading technology innovators for purposes related to training artificial intelligence systems in a manner that respects personal and intellectual property rights. Stability AI did not seek any such license from Getty Images and instead, we believe, chose to ignore viable licensing options and long-standing legal protections in pursuit of their stand-alone commercial interests.”
This case appears on the face of it to be more related to the payment of licensing fees rather than the use of the data. It will be interesting to see how it pans out.
In order to keep this article as short as possible, I have only scraped the surface of the copyright issues with AI generators. It is an area of law where there is a lot of “gray” meaning that there are a lot of things that are open to interpretation due to the way copyright works at creator and societal levels on a case-by-case basis. The cases above may give some clarity to the issues of AI-generated art which may one day be codified in legislation.
Are there any other legal issues?
So far, we have addressed copyright as it relates to the training data and the output of images in one case. There are other copyright issues that we will be looking at from the point of view of creators of AI art further on in this article.
Most attention at the moment is being given to the issue of copyright, but there could be other legal issues that may relate to a particular case. Depending on the jurisdiction, there could be issues arising from tort law – including the right of publicity (right to privacy or unfair competition in other jurisdictions), emotional distress, and defamation. There is also of course potential for crimes to be committed as well including scams, forgeries, and deep fakes.
Can you copyright AI Art you have created?
The short answer is no, ….well usually no. Unlike other areas of copyright law, the fact that a human has to create art has been well-established since the 1880s.
The Copyright Act provides ” s102. (a) Copyright protection subsists, in accordance with this title, in original works of authorship …..”. Explained simply work is “independently created by a human author and has a minimal degree of creativity. Independent creation simply means that you create it yourself, without copying. The Supreme Court has said that to be creative, a work must have a “spark” and “modicum” of creativity.” – Copyright Office USA
Back in 2011 photographer, David Slater traveled to Indonesia to photograph the wildlife in a National Park. While he was there one of the macaque monkeys, Naruto, grabbed his camera and took a selfie. The photograph was later published in the Daily Mail and other publications.
A few years later in 2014, Wikipedia uploaded the image listing it as public domain. Wikipedia asserted that since the monkey took the image, it could not hold copyright and therefore it was public domain imagery.
Naruto v. Slater, No. 16-15469 (9th Cir. 2018)
Throughout a book that Slater and others published in 2014 with the monkey selfies, Slater admits that Naruto took the selfie images “Sulawesi crested black macaque smiles at itself while pressing the shutter button on a camera.”
In 2015 People for the Ethical Treatment of Animals(“PETA”) and another filed a complaint for copyright infringement against Slater and others on behalf of Naruto.
PETA argued the term “authorship” under the Copyright Act was “sufficiently broad so as to permit the protections of the law to extend to any original work, including those created by Naruto.”
Slater argued that as an animal Naruto could not claim copyright “since the Copyright Act did not explicitly give non humans standing to sue for copyright infringement. He called the notion that a monkey was an “author” under the Copyright Act a “farcical journey Dr. Seuss might have written.”
The District Court said in its judgment that “only works created by humans were copyrightable” relying on the Copyright Act and the Compendium. The Compendium is the practical guide to how the Copyright Act works. In this, it states ” works created by nature, animals or plants cannot be registered”. The case went to the Court of Appeal where the copyright aspect was not addressed, rather the issue of whether Naruto and his next friend, PETA, could even bring the proceedings.
Copyright Registration of AI Comic Book – Zarya of the Dawn
In 2022, Kris Kashtanova created a comic book titled Zarya of the Dawn. The text was created by Kris and the images was created by the AI art generator, Midjourney. Kris lodged the comic book at the US copyright office for registration which was approved. The use of AI was disclosed on the cover of the book.
Later, the Copyright Office reviewed the registration and claimed that it was registered as an oversight.
The Copyright Office asked Kashtanova to “provide details of my process to show that there was substantial human involvement in the process of creation of this graphic novel” which she did claiming “a creative, iterative process which she describes as “working with the computer to get closer and closer to what I wanted to express.” This process included multiple rounds of composition, selection, arrangement, cropping, and editing for each image in the Work. Her efforts make her the author of the Work, including authorship of each image in the Work. The computer programs she used including the Midjourney image creation service, were but “an assisting instrument” to Kashtanova.”
The Copyright Office ultimately revoked the registration of the visual works as it did not satisfy the element of “human authorship” required. The text that was created by her though was protected.
Copyright Office – AI Update
On 23 March 2023, the Copyright Office issued a policy update and advised of an inquiry to take place later this year. You can read the Copyright Office Guide on Publishing AI Works here. As part of this update, it stated that “The Office recognizes that AI-generated works implicate other copyright issues not addressed in this statement. It has launched an agency-wide initiative to delve into a wide range of these issues. Among other things, the Office intends to publish a notice of inquiry later this year seeking public input on additional legal and policy topics, including how the law should apply to the use of copyrighted works in AI training” and the resulting treatment of outputs.
The Guide confirms “the term “author,” which is used in both the Constitution and the Copyright Act, excludes non-humans“.
This does not exclude the use of AI completely in creative works however, with the office stating “whether the `work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.” In the case of works containing AI-generated material, the Office will consider whether the AI contributions are the result of “mechanical reproduction” or instead of an author’s “own original mental conception, to which [the author] gave visible form.” The answer will depend on the circumstances, particularly how the AI tool operates and how it was used to create the final work. This is necessarily a case-by-case inquiry.”
At first glance, it seems that the Office is ensuring that authors who reiterate their own works with AI can claim copyright protection but this would depend on the circumstances. The Guide and the Act always operate on a case-by-case basis and don’t cover blanket statements on what is and is not acceptable for registration.
It is important also to understand that this is a guide. The terms of this guide and the refusal of registration for copyright can always be disputed in the courts. In addition, legislation can always be changed and court cases can also add to the body of law when interpreting the legislation. It is not a black and white issue.
Authorship and Creativity – The Future
The thing I’ve always loved about the law is it evolves…… although slowly.
There was a time that photography was not considered “creative” and a photographer could not protect their work through copyright. All that changed in the 1880s when the Courts recognized that a camera is a tool but the photographer is the one that arranges and sets the scene, ensure the lighting is correct, and composes the subject, and composition of the image. All of which require considerable creativity.
At the moment, Copyright law does not consider the act of creating prompts sufficient enough to satisfy the requirements for authorship and creativity. I would not be surprised though, if like photography, one day the process of creating prompts and refining them is enough to satisfy these elements.
If you want to read more about “authorship” in the world of AI here’s an interesting article from the New Zealand Law Journal.
You Create, The Generator Takes
As well as not being able to register your work created with a generator you may have given more than you bargained for to the generator. Make sure to read the Terms of Service carefully. The Midjourney Terms of Service basically state that it is allowed to do anything with the artwork you have created,
“Rights You give to Midjourney
By using the Services, You grant to Midjourney, its successors, and assigns a perpetual, worldwide, non-exclusive, sublicensable no-charge, royalty-free, irrevocable copyright license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense, and distribute text, and image prompts You input into the Services or Assets produced by the service at Your direction. This license survives termination of this Agreement by any party, for any reason.”
If you are a creator this could affect you in a number of ways. If the service distributes your prompts and artwork you could be in effect competing with yourself as others use your prompts or image seeds. The service could also distribute or create further works which may end up devaluing it in the marketplace.
Artists frustrated with the use of AI and alleging that it is stealing their work decided to goad Disney into taking action. Disney has deep pockets and is well known for vigorously protecting their copyrighted works.
Eric Bourdages artist and game developer decided to create images first of Disney characters and then later of Marvel, Nintendo, and DC characters. He was able to successfully create many images using Midjourney proving his point that those copyrighted character images were in the dataset.
He said: “The obvious issue I am opposed to in my thread is the theft of human art,” he said. “People’s craftsmanship, time, effort, and ideas are being taken without their consent and used to create a product that can blend it all together and mimic it to varying degrees.”
He created a number of images using AI Generators to create images of Mickey Mouse and encouraged others to sell the images or use them on products.
Just so we are on the same page. So far we know that if you create artwork using an AI art Generator that:
- generally, you cannot register your copyright and protect the images therefore anyone can use the art you created;
- depending on the terms of service, the AI generator may be able to do what it pleases with your prompts and creations; and
- you own the assets and therefore may be liable for any copyright infringement or other claims
AI Art and Coloring Books
Now that we have gone through the background and have an understanding of the topic, let’s dive in and talk about what really matters, our beautiful coloring books!
Remember at the start of this unbelievably long post I discussed the type of coloring books that are currently available on the market? Here’s a refresher –
There are 4 main types of art that you will find in your coloring books:
- Hand drawn
- Public Domain
- Computer Generated
We can categorize the above list a bit more by sorting the types of books into the level of difficulty and skill.
If we did that we would then find that hand-drawn books are the most time-consuming, and difficult and require the most skill out of the list. Photographic would be my next choice provided that the photograph was also taken by the creator rather than an image site. The others I would assess due to the time and level of skill and difficulty as Quick Coloring Books.
Quick Coloring Books
Since 2015 or so, marketers have been spruiking the benefits of creating coloring books for both children and adults. The books are often seen as “low content” meaning that they take little work to create and can be created easily by marketers and non artists. This means that you can quickly create a lot of books and have them as passive income earners.
As well as requiring little skill to create beyond knowing how to use a couple of computer programs they tend to be easy to create. They are also low-cost to create because other than buying software and using a word processing program nothing else is required.
The books can be listed on Amazon (originally Createspace) and now KDP where they are only printed when a customer orders the book which is known as Print on Demand (POD). This means that there is no inventory to hold, store, and move and very few fees until a sale is made. Once a sale is made after 30 days or so (depending on the mood of Amazon haha) payment is made to the creator.
The reason that they are considered “quick” to create is that compared to writing a novel or nonfiction work they can be created using public domain images, clip art, desaturating photographs, or using software to create patterns, mandalas, or kaleidoscopes.
These are not artist-drawn books that take months to draw, compile and create but low-effort books when compared to hand-drawn artist-created books.
AI- The New Player
This is where AI generators change the playing field completely.
Since AI generators became widely available to the general public there has been a resurgence not just by marketers who initially pushed the Quick Coloring Books but by a wide range of people; creators, YouTubers, and side hustle mentors. You name it – creating AI coloring books is huge right now.
Here’s just a selection on YouTube at the moment:
If you were creating Quick Coloring Books you are now competing in a market of art that can be created quicker and often better than what you have been creating so far.
Artists that create hand-drawn work, now have serious competition as AI Art can go head to head with hand-drawn art because when it is done well, it can be incredibly realistic. Realistically, for hand-drawn artists, it might look like they are competing for their slice of the coloring book market with anyone that can create (or buy!) a prompt for use at one of the many free or paid AI Generators. Not only that, as the AI generators can create more illustrations in the time it takes an artist to sharpen their pencil it might feel like AI coloring books will take a larger share of the volume in the coloring marketplace. In my opinion, neither of these things is likely to happen to any great extent.
Coloring Book Market
The heyday for adult coloring books was way back in 2014-2016.
Since that time many artists that created coloring books in that first wave of adult coloring love have now left the market and Time Magazine famously said the market was in decline.
A new wave of adult coloring happened during the pandemic when many people were in lockdown and discovered a new hobby. Some have stayed with the hobby and again some have left.
AI art generators are marketed toward non-artists. People like me. Anyone can use them so the barrier to entry is low.
Sadly, most of the people that attempt to enter the coloring book market are not going to achieve their goals. I say that based on my knowledge of this market. It is not a get-rich-quick business. In fact, most of the artists you know and love do not make a full-time living from the sale of their coloring books. It is only a small percentage of artists that can afford to live off the royalties or income generated from their artwork.
Even though AI-created coloring books can be created faster, it does not mean that there are more buyers in the market to purchase the books. As you can see from Google Trends image the market is declining. Add to that there is currently a decline in disposal income meaning people have less discretionary spending for hobbies.
A declining market means that there will be greater competition with existing players to retain and build on their existing fan base. If there is a glut of coloring books on the market there will not be enough consumers to support an expanding marketplace. Some artists and creators will survive – some won’t.
AI Coloring Books
I’ve noticed there have been a lot of coloring books released lately that are obviously created with AI Generators. Many of them—though not all—include this information in the title.
A quick search of Amazon for “AI adult coloring books” and related keywords, reveals hundreds of listings.
As AI art is still in its infancy, it’s far from perfect and currently easy to identify. From a distance, these books look quite good and creative. Look closely. AI Generators have a habit of making the same mistakes over and over again. The usual things that signify AI art that hasn’t been edited out are things like odd-shaped eyes, too many teeth, too many legs, arms, and especially fingers – the six-finger person seems to feature a lot!
The image used as the feature in this post took over 6 variations for me to create as the woman had multiple arms when generated. In the background, you will often see odd shapes, curls, and random “bits” or artifacts. Sometimes, you will see a bit of a watermark or signature on the image as well. Look at the small details. Look closely at the picture and envisage how you would color it. Many of the quirks in AI art can make it frustrating to distinguish what an element is and as a consequence how to color it.
New upgrades on AI generators over the last few weeks have been significant. With the release of Midjourney 5 more realistic images will be created and many of the problems and quirks in AI art will gradually be a thing of the past with each update.
A search of Etsy Marketplace also reveals pages and pages of AI-generated coloring pages. There are others that have been AI generated as well but they have not been labeled as such. Some have not been modified which appears to be in breach of Etsy’s policies.
There is some AI art that is being produced in the coloring book world that does show care in their creations. Many of the pictures are imaginative but also colorable. They have taken time to polish their AI Art to correct imperfections, edit out artifacts and generally clean the picture making it fit for publication.
Artists using AI – Hybrid Artists
As well as solely AI artists and solely hand-drawn artists there is now a new breed of artists which uses both – Hybrid Artists. Some people call them collaborative artists because the artist works with technology to create art as well.
How Artists are Using AI
1. Idea Generation: Use AI art generators to create a variety of initial concepts or sketches. This can help explore different ideas and styles before committing to a specific direction for your illustration.
Example – Cueva Gallery contacted Mas, an Italian self-taught painter that works with a variety of mediums, and had her take part in an experiment on incorporating AI in her creative process. Mas initially struggled with the process of using technology but eventually found that the way it suited her work best was as an inspiration. Using pictures produced by AI on endangered species she found that “The stimulation of the imagination due to the fact that the residual images are blurred could serve very well as abstract paintings. I did not expect at the beginning to be able to find a way to look at images produced by AI and incorporate them into my practice. But in the end, I did it in my own way.” — Mas.
2. Color Palette Selection: AI art generators can suggest color palettes based on your input or reference images that you upload. This can help choose harmonious color schemes for colored inspirational images and cover art as well as in the marketing of the coloring book
3. Background Creation: Creating backgrounds or scenery with AI which can be incorporated into the illustration, freeing up time to be spent on the focal point of the illustration
4. Style Transfer: Upload your own artwork or concepts and apply a different technique or art movement to the illustration with AI. This allows an artist to experiment with different techniques before deciding whether to commit to drawing the illustration.
5. Enhancement: Refine your artwork by uploading your images and experimenting with light, shade, and details
6. Collaboration: Use technology and the output of AI art generators as a starting point or reference for illustrations. Combined with experience and knowledge of art the final artwork becomes a collaborative piece of art.
7. Recreation: Many artists are using AI to recreate their own work. Gavin Goodman, an advertising photographer, stirred debate by “stealing” his own images and using artificial intelligence to recreate them.
He said he was able to produce new images faster, economically, and more creatively than originally.
Goodman was so impressed with the results that he set up his own studio offering brands and agencies AI artwork and AI plus photography-generated imagery.
Finland-based artist Antti Karppinen also has spent time recreating his own photographs.
“In one year, AI developed so much,” Karppinen explains to My Modern Met. “I started doing AI images last year, and have seen it evolve into something rather amazing. I just wanted to see what would be the quality of the images if I input my images as a reference and let the AI do its thing with a prompt.”
Photographer Jeff Hayward used Midjourney to reimagine his own photographs as well. He used an upload of his images and a text prompt to produce new but very similar images.
He was happy with the results stating “I maintain it takes creativity and know-how to truly stand out as an artist, whether you’re writing, drawing, taking pictures, or making AI images. It takes imagination to dream up concepts that the tools translate for you. AI is a major leap, but it’s really just the next step of art’s evolution.”
Although the examples I have given are all photography examples these can all be applied to the coloring book world of illustrations.
In the coloring book world, Nori Art Coloring on Etsy has a range of hand-drawn pages on her store and pages that she states were created with the assistance of AI. The AI-generated pages are of high quality and the images have been edited and cleaned up to ensure the best coloring experience for buyers. Nori Art also offers hand-drawn artwork. Her artistic background is reflected in the high quality and “colorability” of her AI-generated artwork.
Many artists are using their skills and knowledge of art, combined with AI on freelancing sites. The freelancing site, Fiverr revealed that artists with skills are in hot demand ” Fiverr’s new Business Trends Index reveals hirers are paying between $600 and $1200 a project for art created using artificial intelligence tools such as Midjourney, Dall-E 2 and Leonardo.AI.” Fiverr community manager Olly Woolrych said there had been a “huge boom” in demand for AI art skills “Some of the freelancers have a background in illustration as an artist, others are wordsmiths and creative writers … it’s a real blend of different skill sets according to the Daily Telegraph (an Australian newspaper)
How do Colorists Feel About AI Coloring Books?
While I have not conducted any surveys on this question, I do keep an eye out on the coloring community and take note of how people are reacting to AI Coloring Books. It seems that there are really three camps. One that will choose to support hand-drawn artists, others who are excited and intrigued by the images produced by AI and want to color them, and a smaller group who are happy to color from either camp, as long as the picture is appealing.
Saying that I have noticed that many colorists that have purchased some AI Coloring Books have been disappointed and would not purchase them again. The two main issues are:
- they are difficult to color due to the unidentifiable fragments of an image and quirks on AI art that have not been carefully edited or cleaned; or
- they feel deceived that the book was not advertised as AI and they would not have purchased the book had it been disclosed.
How do Handdrawn Artists Feel About AI Coloring Books?
In the current environment with the sudden influx of AI-generated coloring books, many hand-drawn artists are feeling the effect. On a personal level, many are vocal that artists’ images have been used without their consent in image data sets and to train AI Generators. Effectively this usage and technology allow anyone in the world to become a creator that will eventually produce artist-worthy quality imagery. The fact that this will be done for a small fee rather than years and years of training, studying, and practice and in the time it takes to sharpen a pencil just adds insult to injury.
Add to that the ongoing discussion about AI art coloring books seems to have had a negative effect on many artists who may feel dragged down and depressed by the various points of view on this emotionally taxing topic. Some have experienced negativity from members of the coloring community for sharing their opinions on AI art.
As well as the personal effects felt by many artists there is also the knock-on financial effect. Many have reported a drop in sales. It is not known though whether this drop in sales is a reflection of the current economy or the introduction of a new competitor in the coloring marketplace.
How can Handdrawn Artists Compete in the Changing Coloring Marketplace
In my opinion, hand-drawn artists are in the strongest position in the coloring marketplace as compared to Quick Coloring Books and AI-Generated Creators.
Handdrawn artists, in my opinion, will become more niche in the coloring marketplace. In a sea of Quick Coloring Books and AI-Generated Books, the hand-drawn aspect will become a standout.
Looking at history we have seen the appreciation of traditional arts as new technologies emerge. Below are some examples –
Photography: When photography was invented in the 19th century, it disrupted the traditional art world by offering a more realistic and accurate representation of the world. Many thought that this would see the end of the painting. That has not happened. Instead, painting grew while artists explored different styles. Original paintings from the pre-photography era became more valuable and sought after as they represented a unique perspective on the world that was no longer possible to capture.
Printing press: The invention of the printing press in the 15th century made it possible to mass-produce books and images. However, original woodcuts, engravings, and etchings from before the invention of the printing press became more valuable as they represented a time when every copy had to be made by hand.
Film: When the film was first introduced in the late 19th century, it was seen as a novelty and not taken seriously as an art form. As filmmakers began to experiment with the technology films became more sophisticated and complex. This led to a renewed interest in early silent films and original movie posters, which became highly collectible.
Over the years the art world has changed. New technology has been introduced that most hand-drawn artists who create coloring books use – things like Procreate, Photoshop, and Adobe Illustrator to name a few. This digital technology has allowed artists to create coloring books and cover art that is refined, edited, and colorable.
As AI Generated art develops and perhaps increases in popularity, it is my opinion, that traditional hand-drawn art will become more desirable, valuable, and collectible due to the unique “human” element of the creation.
This inexplicable and difficult-to-define “human” element is what will set hand-drawn artists apart from AI creators. Artists who are not using AI can still compete by focusing on what makes their work unique and human.
Here are a few ways hand-drawn artists can differentiate themselves and compete with AI-generated art:
- Do Not Compete: That would be like apples competing with apples. Handdrawn artists need to focus on being oranges, in my opinion.
- The Human Element: One of the key differences between traditional art and AI-generated art is the human element. Showcase the skills, techniques, and personal style that make your work unique.
- Highlight Process: The process of creating art can be just as important as the final product. Artists can share their creative process with audiences, showcasing the effort and thought that goes into each piece.
- Collaborate: Engage with fans and ask for their feedback, and thoughts on coloring book ideas and other products they would like to see you create and sell
- Storytelling: Art has the power to tell stories and evoke emotions in ways that AI-generated art may not be able to replicate. Traditional artists can use their work to tell stories and communicate ideas and inspiration that are personal to them.
- Experiences: Artists can offer workshops, demonstrations, and interactive experiences that bring their work to life. This is an opportunity to further engage with fans, collaborate and build a larger audience
My Recommendations for Artists and Publishers
- Check to see if your artwork has been added to image data banks. You can do this by going to Have I Been Trained a free website created by artists. Follow the instructions on the website. If you do not wish to be included in the data set for Stability AI you can sign in and request an opt-out of your artwork, which Stability AI has agreed to honor. Using this opt-out system over 80 million artworks have been removed from the data set in the last few months.
However, there are problems with opting out mainly:
- Stability AI is not the only AI Art Generator system being used. More are being created constantly and many use their own created data sets; and
- the time associated with uploading, locating, and checking a large portfolio of work on Have I Been Trained is incredibly time-consuming and takes away from paid work
Unless the law is changed to allow artists to Opt-In (rather than the reactive opt-out) this will be an ongoing and frustrating game of whack-a-mole.
Deviant Art and Art Station have changed their systems to ensure artists can opt out of training data sets. On Art Station images need to be tagged with #noAI and on Deviant Art all work is automatically opted out of the training data sets. This is not a guarantee though that artwork will not find its way into a data set by other means.
- Keep updated with the Copyright Office’s proposed public consultation later on in the year. The Office intends to publish a notice of inquiry later this year seeking public input on additional legal and policy topics, including how the law should apply to the use of copyrighted works in AI training” and the resulting treatment of outputs. This is an opportunity to be heard by lawmakers on this issue.
My Recommendations for AI Creators
You may not agree with me and that’s fine :). These are the things that I would like to see happen in the coloring book world. Although, this article is really about the impact of AI art generators in the coloring book world it also applies to many other compilation and Quick Coloring Books as well.
Transparent: Disclose clearly on the cover, title, and description of your artwork if the image has been created with AI. This not only builds trust with the coloring community but ensures that consumers do not feel deceived or misled when making a purchase and can make an informed buying decision
Correct: Take the time to clean up the images, removing artifacts, rasterized line art, and AI quirks to ensure that every part of the page is not only identifiable but also colorable
Style: Use creativity to develop your own style of artwork that is unique to you rather than referencing a particular artist’s style
Ethical AI: Consider using generators like Firefly by Adobe that use images in its dataset that have been licensed or in the public domain or use your own images, photographs, or public domain art as a base. This could become even more important at a later stage, depending on the results of current cases before the courts
My Recommendations for Consumers
Investigate: Before investing in a new coloring book be sure and check out the illustrations and see if they are to your liking. There are a number of ways that you can do that – the “look inside” feature on Amazon which often shows a few pages, video uploads from purchasers on the Amazon product page, the artist’s website/Facebook/Instagram page, asking in Facebook coloring groups, watching flip throughs by independent reviewers, and YouTubers. If the artist is not known to you, check out their social media, website /author profile on Amazon, or About page on Etsy. This will help you understand the way they create their coloring book pages and give you ideas of their style and if it resonates with you.
Be Kind: AI-Generated art has polarized many members of the coloring community. I have seen many instances of bullying, intimidation, and humiliation of people as a result of their opinion of AI Generated Art. This is deeply concerning. Coloring communities should be safe spaces where individuals can unwind, appreciate art, and reap the therapeutic benefits of coloring. It’s fine to have your own point of view but it’s crucial to convey it in a respectful and considerate way that acknowledges others’ views. As the saying goes, you don’t need to set yourself on fire to keep others warm.
Support: No matter what style of art you like to color, it’s always important to support the creator. Continued support can mean a lot, especially during difficult times, to ensure that artists and creators are encouraged to continue to create beautiful artwork that we can enjoy, color, and collect. This can be as simple as liking a post, making a comment, or leaving a review.
If you made it this far, thank you. You have earned a cup of coffee, chocolate, and a nap!
I hope you have gained some insights into the world of AI art and the current issues affecting the coloring community. As the author of this blog, I do not intend to influence your coloring preferences or urge you to support specific artists or creators. Instead, I believe that it’s important for consumers to assess the value of a product and make informed decisions before making a purchase.
For me, the value of a coloring book can mean different things, including the artwork’s monetary worth, the creative process, the experience I would get from coloring it, the appeal of the artwork, and my relationship with the creator.
I am not the “Coloring Police” and recognize that everyone has different preferences when it comes to coloring. While I believe there is room for both hand-drawn and AI-generated art in the coloring book world, the question for consumers boils down to where they see the most value for their time and money.
As we navigate these evolving times, it’s uncertain where this will all end. My hope is that this article has provided you with a fair and balanced view of the issues to help inform your decision-making before your next coloring book splurge.
I would love to know your thoughts in the comments below.
Happy creating & coloring x