The future is not looking good for stock photography and illustration or for the photographers and artists who supplement their income with it. After spending some time with Midjourney and DALL-E 2, I can feel the shift. The day is fast approaching, and the ability to generate a unique high-quality image of just about anything you wish to conceptualize will be possible. In some cases, the day has already arrived. Of course, every model will have its strengths and weaknesses, but each will do far more amazing things than the specific examples I’m using here.
(I should note, I’m only focusing on two of the tools I’ve been playing with. There are far more. Artbreeder, Deep Dream Generator (from Google), Wombo Dream, and NightCafe are just a few of the AI image generation tools that are producing interesting results.)
I began thinking about this as a result of a conversation in our living room with my neighbors. We played a game where I would show them Midjourney outputs, and they would try to guess the image prompt. It was surprisingly enjoyable. When I showed them one image, a neighbor said, “that looks like a photo you might have taken in your creek in West Virginia.” And that was indeed the prompt. It was confusingly realistic as it appeared on my TV screen. That sense of reality caused them to ask, “will we even need to take nature photos anymore?”
It was an interesting question, and my personal bias and love of photography caused me to quickly answer, “yes, of course.” But it did force me to stop and wonder. Many weeks later, this post is the result of that wondering. So let’s use the West Virginia creek example for our focal point.
Stock Photography: WV Creek
All of the images below were created with the following prompt in DALL-E 2: “West Virginia creek surrounded by lush forest, autumn, fall colors, thick woods, tall trees, low-angle perspective, golden hour (or morning fog), cinematic atmosphere.”
The examples above are fairly low-resolution images by today’s digital standards. The generated files are 1024×1024 pixels, and each set of four images took less than one minute to generate. The DALL-E 2 beta does not currently offer a way to specify different levels of quality or aspect ratio like Midjourney does. As of this writing, neither allow significantly higher resolutions with their models. But you can see the potential even in this current state. I could easily fool you with these images.
In my testing, DALL-E 2 seems to perform better in generating “real world” images, and Midjourney knocks it out of the park when you are looking for something more conceptual. Below is an example of what Midjourney returned for the prompt: “West Virginia creek surrounded by lush forest, autumn, fall colors, thick woods, tall trees, low-angle perspective, morning fog, high-resolution digital photograph, cinematic atmosphere, hyper-realistic, octane render, trending on artstation.”
To my eye, the images that come out of Midjourney have a more rendered feel, even when attempting to recreate reality. The images from DALL-E 2 have a more photographic feel in general. This carries over into conceptual art areas. I’m going to show a few examples to illustrate this point. DALL-E 2 leans toward better results when you’re attempting to create something you want to look “real,” and Midjourney excels at “concept.” In the “Monet” examples below, DALL-E 2 does a better job interpreting “creek,” and Midjourney seems to excel at the “art.” But here’s the thing, when I’m looking for stock images, sometimes I want real, and sometimes I want out of the box.
You can see similar results when you feed both the prompt: “a West Virginia creek surrounded by lush forest, autumn, fall colors, thick woods, tall trees, low-angle perspective, golden hour, cinematic atmosphere:: in the Disney animation style:: 2d flat colors, vibrant, vector art.” The prompts, for the record, need to vary slightly from one platform to the other because of the way they handle styles, but they are essentially the same. The results, however, are vastly different.
We Are Only At The Beginning
We are quite literally in the infancy of these technologies. Already, however, the results are astounding. And when you go even more conceptual, things ever very interesting very fast. For example, I was watching a documentary, and a man was wearing a t-shirt with a “Día de Muertos” skull design. Of course, my brain instantly wondered how these two AI platforms might treat that, so I fed the following prompt into each: “Día de Muertos skull wearing a jeweled crown, octane tender, cinematic lighting.” Again, the Midjourney results feel more conceptual, and the DALL-E 2 results trend toward realism.
Stock Photography: Día de Muertos – Midjourney
Stock Photography: Día de Muertos – DALL-E 2
Stock Photography: The Future Is Synthetic
The world of stock photography is about to change dramatically. With the advent of AI systems that can create realistic images and art from a description in natural language, we can say goodbye to much of the traditional stock photo industry.
Traditional stock photo companies rely on human photographers to capture specific, pre-determined images. But as text-prompted AI-generated images improve, any scene or object will be able to be created on demand, without the need for a human photographer. Businesses and individuals will be able to get exactly the images they need, without having to go through a middleman. I can already use any image I generate from either service for commercial purposes. I can resell them if I wish.
So what does this mean for the future of stock photography? This shift will have a major impact on the stock photography industry. Will it replace custom illustration and photography needs? No. But the demand for low-cost, high-quality digital media is beyond what artists are able or willing to deliver. These tools will fill that gap, perhaps even create it.
With the ever-increasing power of AI, we can expect these platforms to become increasingly better at creating realistic images and art. Stock photo agencies will have to find new ways to make money, and many photographers will lose another stream of revenue. They will be replaced by “synthographers” creating synthetic media.
It’s going to happen. And faster than you may be ready to believe.
[…] wrote about the future of stock photography on August 13, 2022, and the examples in that post now seem quaint. Take a look at the examples in […]