AI can undoubtedly create stunning visuals, but it also raises some valid concerns about originality and copyright. Addressing these concerns has become essential in the world of art and fine-art photography in particular.
In this article, I will focus on how to maintain originality when using AI imaging platforms by using your own photographs when generating images. Doing so not only adds a personal touch, but it also gives you more control over the creative process while helping with important legal considerations. Let's get started by reviewing some tips.
8 Tips To Use AI Imaging and Maintain Originality
Avoiding Overreliance on AI: It's important not to rely too heavily on AI for every aspect of image creation. Incorporating AI as a tool to enhance your existing images and your creative process rather than replace it can help maintain originality.
Human-AI collaboration: Use your own photographs as the basis for your AI-generated images. You can think of this as human-AI collaboration. In this way, you can combine your personal creativity with the capabilities of AI. This “collaborative” approach can result in unique and innovative art that wouldn't be possible with AI alone. More on this below in the section “Our Approach To Using AI Image Generators.”
Focus on customization over AI creation: By tailoring an AI-generated component or enhancement to your existing image, you can help ensure that the final result will meet the test for originality (i.e. the AI-generated elements are selected or modified "in a sufficiently creative way." See our blog post "Ethical and Copyright Considerations of AI-Generated Images In Photography".)
Legal Frameworks: Be aware of and stay informed about copyright and intellectual property laws. Ensure that the AI tools and datasets you use are in compliance with these evolving rules and regulations. This can help protect your creations and maintain their originality. This isn’t always easy but should improve in the coming months (hopefully not years).
Ethical Considerations: Address the ethical aspects of your AI-generated art. For example, if you do people photography be transparent with your customers about your use of AI in the creative process.
Mixing and Remixing: Experiment with mixing and remixing AI-generated elements with traditional artistic techniques. This can lead to truly innovative and original results that use the technology to further your human expression, not replace it.
Constant Learning and Adaptation: The field of AI is continuously and rapidly evolving. Stay updated with the latest advancements and adapt your creative process accordingly. This can help you stay ahead in creating original and cutting-edge art.
Feedback and Critique: Encourage feedback and critique from peers, clients, and art experts. This external input can help refine the art and ensure its originality.
By addressing these concerns and adopting a thoughtful, client-centered approach, you can confidently use AI to create original and meaningful photographic art.
Our Approach To Using AI Image Generators
We have used several popular AI tools but have recently settled on two for most of our work: Midjourney and Adobe Firefly. The latter is the foundation for other Adobe products like Photoshop Generative Fill. Which works best for you depends on your images and composition goals. We continuously experiment with both. The Midjourney image generator is also one of the few, at the current time, that allows a 4x upscale to over 4000x4000ppi. We have concerns about their training datasets, which is another reason we always use our images as the basis of our digital artwork.
Choosing the Right Photograph
What makes a good reference image to upload into the tool?
To create stunning AI-generated images, I recommend getting started by selecting a high-quality photograph with clear details. However, remember that fine details can be altered in surprising and not always desirable ways with distortion, noise, blur, and weird juxtapositions to varying degrees. Here are a few other tips that may help, particularly when getting started.
Opt for vibrant colors and interesting, simple compositions to use as a reference image. We sometimes edit an existing photograph to simplify the subject matter and backgrounds before uploading it as a reference image to a tool like Midjourney.
Good focus, lighting, and minimal noise will help ensure better results.
Experiment with different types of images, such as landscapes, still life, or portraits, to learn how the platform reacts to your images.
Prepare your image for the AI platform. This usually means sizing it somewhere under 2500px in the longest direction, perhaps exporting a layer, and making sure your image is well within the histogram limits… we call this providing headroom for the image generator and additional processing.
Depending on the platform, you can upload multiple reference images that the generator will blend into a single generated image. Midjourney will accept up to five images to blend. For example, you could upload both a flower and creek image, instructing the platform to use the creek in the background. The results and possibilities are limitless. The important point is the primary elements are based on your own creativity.
After uploading the reference image(s), write your prompts with as much detail as possible. Checking out the prompt support pages on your desired platform is a good idea. It does take practice, so be patient and keep trying.
In Midjourney you can set the weight of the reference image. This tells Midjourney how closely you want the final result to reflect your image. The range for this command is "-- iw 0" to "--iw 2.0" where the latter will most closely resemble your reference image. You can check out all of the: Midjourney Parameters.
It is common to have to go through several iterations and variations to get a generated image you feel meets your needs and is suitable for additional customization. Depending on the platform, you may be able to upscale the image before downloading it for further editing.
How To Use Your Own Photographs With AI
AI can be used like any other imaging technology to enhance or modify an existing photograph you create, particularly in combination with Photoshop, Topaz Photo AI, On1 Photo Raw, filters, textures, backgrounds, layers, and other image editing tools in creating an image. In this way we think of and use AI as just another tool in our toolbox.
Our AI Workflow
As stated above, we use AI in combination with our original images and other technologies using some form of the following workflow. Your workflow may differ, of course.
We always begin with one of our original photographs that we prepare in Photoshop for importing to AI. Sometimes, we’ll make and export a layer for the AI component, such as just the subject or background. We also modify the metadata to include the copyright and remove any info we do not want to be ingested by the platform. (See the section “What Does the AI Platform Do With Your Uploaded Photographs?” below).
We upload the prepared image into Adobe Firefly or Midjourney to be used as a reference image.
We reimagine this reference image through what usually turns out to be a series of successive prompts and variations that we refine until we get something close to our vision for the image or element.
Once we generate a successful candidate, we upscale it by 4x (Midjourney).
Finally, we import or open the resulting image into Photoshop, often as a layer, where we continue the customization further with a blend of textures, filters, backgrounds, brushes, masks, etc., using all of the imaging options and skills in our toolkit. We will also use applications such as On1 Photo Raw, Topaz Photo AI, Topaz Studio, and Nik, as needed on almost every image, as each has features that complement Photoshop for customizing and preparing our final master image.
The critical factor with respect to AI is that the generated image's primary basis is one or more of our original photographs, not the AI tool's training dataset alone.
By combining AI technology with your photographs and further editing, you can create a stunning, personalized, original image.
What Does the AI Platform Do With Your Uploaded Photographs?
There’s always a downside, right? Keep in mind that most AI platforms will, by necessity, so they say, ingest your image into their training dataset. This raises a significant privacy issue due to the potential to access and utilize whatever is in your metadata. Uploading your photos to an AI platform effectively grants the software access to the image and possibly a range of personal data, location details, description information, and various other metadata values. If you photograph portraits or people, you should also be aware and considerate of your subjects’ privacy before uploading their likenesses or any metadata containing info about them into an AI platform. For these reasons, we strongly recommend being transparent with your customers and modifying the image’s metadata in the file preparation stage before uploading it to an AI platform.
The Need for "Do Not Train"
In my opinion, the AI platforms should not ingest your uploaded images into their training datasets. At the very least, the issue can be easily resolved with the addition of an option called “Do not train.” For example, Adobe plans to add a “Do not train” tag to Firefly and its Generative AI tools. Other platforms should as well. Artists and the creative industries need to advocate for the addition of this essential feature and support those platforms that respect artist rights in this way. Besides, it would be so easy for the AI platforms to do so! You can join the Content Authenticity Initiative to stay informed on this topic. It's free. In the meantime, each artist needs to decide on the pros and cons of whether to use their own photographs when generating AI images.
It is possible to create original work harnessing the power of AI using your own photographs. Doing so opens up a world of creative possibilities. With the right original images, tools, and techniques, you can elevate your photography and create visually captivating original images that leave a lasting impression.
Note: Aiello Studios does not earn revenue from links or companies referenced in this post.