Comparing Today's Leading AI Image Generators

⚠️ TRIGGER WARNING: If discussions about AI-generated imagery and its impact on photography and visual arts bother you, you might want to SKIP THIS ARTICLE!

Hey there, fellow image-makers. I’ve been thinking a lot about this strange new world we’re inhabiting together. Photography is—and always will be—about catching light on film or sensor. It’s still about waking before dawn to catch that perfect moment, fingers freezing on the tripod as we wait for the light to break just right. Those experiences—the cold mornings, the missed shots, the unexpected victories—they’re the essence of what we do as photographers, and no algorithm will ever change that.

This week witnessed another significant advancement in AI image generation with both Gemini and ChatGPT releasing new tools. As a long-time Midjourney user, I’ve observed its evolution into what I considered the leader for photographic realism—but the landscape is shifting rapidly.

Though we’re a photography publication, it would be negligent not to address the transformative impact of AI image generation on our field. This technology is fundamentally altering the industry, leaving many commercial photographers and graphic designers searching for new paths. The discourse about the loss of human connection and photographic soul runs deep, but something more concerning is happening beneath the surface.

What’s being lost in these systems isn’t just the human touch—it’s the nuance of emotional and political expression. These AI generators are built with guardrails that inevitably reflect current societal norms and corporate risk management rather than the full spectrum of human experience. Critical thought and challenging perspectives are being subtly filtered out, homogenizing our visual language in ways we’ve barely begun to reckon with.

My intent here isn’t to philosophize but to examine the practical capabilities of these tools. Today, we’re looking at ChatGPT, Gemini 2.0 Flash Experimental {Image}, and Midjourney to determine their respective strengths and limitations.

In my testing, ChatGPT demonstrates remarkable versatility—its ability to integrate text and manipulate images is so advanced that I’m genuinely questioning my Adobe subscription. When a free tool can accomplish in seconds what would take me minutes or hours in Photoshop, the calculus of creative economics changes dramatically. Gemini excels in speed and text handling, while adding a small watermark in the bottom corner of each image—I’m not yet clear if there’s a paywall or way to remove this. Meanwhile, Midjourney still captures something closer to the essence of photography, despite falling behind in other aspects.

This exploration began as personal research but evolved into something I felt compelled to share with fellow visual storytellers. I find myself simultaneously disheartened and invigorated by these new creative capabilities—a contradiction I suspect many of you share.

Let’s examine what these image generators are truly capable of producing.

REAL OBJECT PROMPT:  Please use this image of the man and have him holding this camera – make photorealistic an that he’s standing on top of a mountain catching the sun going down at golden hour, he’s facing the camera, shot with a Hasselblad at 2.8 aperture, Have the text read “the new Fuji GFX 100RF”.

Midjourney came up with these ridiculous images:

Chat GPT:

It did a pretty good job of following the prompt and the fingers look great.  The camera and the skin looks a little off from the picture.  But not bad.  The text reads correctly which is a HUGE deal. 

Gemini:

It did a pretty good job (except if gave me a little bit of a belly and too many “i’s” and “X’s”.

Let’s test a couple of more prompts and compare:

Golden Hour Landscape Photography: A winding mountain road cutting through autumn foliage, captured at golden hour with long shadows. Dramatic side lighting reveals texture in the landscape while maintaining shadow detail. Shot with a medium format camera at f/8, exhibiting natural depth of field and dynamic range.

Midjourney:

ChatGPT:

Gemini:

Commercial Product Photography: A crystal drinking glass with condensation droplets on a dark surface, shot with dramatic rim lighting to highlight the liquid’s refraction and texture. Perfect reflections in the surface below, creating a high-end product advertisement aesthetic with studio precision.

Midjourney:

CHAT GPT: 

Chat GPT failed at first attempt. And the second.  The third time got a complete image.

What’s unique about ChatGPT’s new image creator is the ability to edit and change parts of the photograph – including adding text.   It did have a difficult time rendering it taking many times to complete.  Eventually I had to redo ask as a new prompt – but it completed it beautifully.

To note, first time did not render. Asked again.  Second time came close.  Could not complete until

Notice how it introduced a highlight at the bottom of the glass.  A nice touch

Gemini:

Remove Power Lines:  “Remove the power lines from this image”

Gemini:

The rendering was really fast as I still wait on the Chat GPT’s output.  It also took out the crane.

Chat GPT:  Did a great job taking out the power lines, and enhancing the image. The detail was improved and the horizon corrected. 

Midjourney:

Midjourney failed at this completely.  It did give some expressive images though but nope.

Sky Replacement:   This is an image that was used when I went to Sedona testing out the Fuji GFX 100S.  Prompt:  “Replace the sky with the milky way.”

Gemini:

Chat GPT:

First pass (it cropped image).  I asked for it do it again at a wider aspect ratio. 

Midjourney:  Not Quite.

Street Photography Decisive Moment: A busy rain-soaked street corner in Tokyo at dusk, neon signs reflecting in puddles. A single figure with an umbrella crossing the street, captured mid-stride with a fast shutter speed that freezes the falling raindrops.

Gemini:

Midjourney:

Chat GPT:

I don’t know if the text is accurate. But, it did seem to follow the prompt well with the raindrops being stopped in air like a fast shutter.

Macro Nature Photography: An extreme close-up of a dewdrop on a spider web at sunrise, with the landscape visible and refracted within each water droplet. Natural bokeh in the background, showing the technical challenges of depth of field in macro photography.

Gemini:

Midjourney:

ChatGPT:

Environmental Portrait: A weathered lighthouse keeper standing in the doorway of their lighthouse, half in shadow, half illuminated by warm interior light contrasting with the cool blue twilight outside. Subtle human expression that conveys years of solitude and dedication

Gemini:

It seems photographically accurate, including the silhouette,  but lacks the emotional human expression.

Midjourney:

ChatGpt:

First time nothing. Asked again to finish image.  And again. Then ChatGPT came up with this:

The face looked really accurate.  I asked it for a close up of the same person but even closer. “great – can you crop in to this into more of a close up? I’d like to experience and feel more of his face and soul”.  It went into “Deep Research” mode which is a fascinating way to see how it comes to the outcome of the image.  You see where it’s sourcing some references in real time (wikipedia, pexels.com, unsplash, etc) 

This is the image it came up with after that:

Typography in Landscape: The word “WONDER” appearing as giant three-dimensional letters nestled within a dramatic mountain landscape. The letters should be physically integrated with the environment—partially submerged in a lake, emerging from forest, or carved from stone—not simply superimposed. Morning mist surrounds the scene with directional light creating realistic shadows that follow the contours of both text and landscape.

Gemini:

Midjourney:

ChatGPT:

Stock Photography Business Concept: A minimalist workspace with a laptop, coffee cup, and small plant on a clean white desk beside a window with soft natural light. Balanced negative space for text overlay, neutral color palette with a single accent color, and shallow depth of field focusing on the laptop screen.

Gemini:

Midjourney:

A pleasant rendition, several choices were inaccurate though like the ceramic handle on the side of a paper cup.  I do like how Midjourney gives different options that you can vary from there. 

Chat GPT:

For fun, I asked ChatGPT to incorporate the Luminous Landscape logo and put it inside the computer on top of a picture from a beach.  It first created the image of the beach.  This is what it came up with:

Ok – so maybe photoshop’s still needed!

Human Emotion Portrait Prompts

These prompts I wanted to really try and see which program would be able to express emotion and real world photographic directions.

The Weight of Memory: Create a portrait of an elderly person sitting alone in their living room, looking at an old photograph held in weathered hands. Their expression should capture the complex emotions of nostalgia, loss, and quiet resilience. The scene should be lit using a single large softbox (Profoto 5-foot RFi Octa) positioned at 45° camera left as the key light at f/5.6, creating gentle shadows that accentuate the facial contours and the texture of their skin. A silver reflector at camera right provides subtle fill. The room is dimly lit with warm practical lights in the background creating bokeh at f/2.8. Shot on a full-frame Canon camera with an 85mm lens at ISO 400, 1/60s. Natural vignetting should draw the viewer’s eye to the subject’s face while a shallow depth of field keeps focus on the eyes and the photograph they’re holding. The color grading should employ a desaturated palette with preserved skin tones, emphasizing the contrast between the vibrant memories contained in the photograph and the quiet present moment.

Gemini:

Chat GPT:

Midjourney:

Some of the renditions are a little scary with fingers and strange things in the holding of the photograph.  It was expressive.

Pure Elation Captured: Create a portrait of a young child experiencing unbridled joy while running through a sunlit sprinkler on a summer afternoon. Their face should radiate genuine delight—eyes crinkled, mouth open in mid-laugh, water droplets suspended in the air around them. Shoot with a Canon EOS R5 and RF 70-200mm f/2.8 lens at 135mm, f/4, 1/2000s to freeze the water droplets, ISO 400. Light with golden hour sunlight as backlight (rim lighting the subject and water spray) positioned at 5:30pm in summer, creating a natural halo effect. Add a large California Sunbounce Pro reflector with zebra fabric (silver/white) at camera left for fill light on the face. A second assistant holds a 1/2 CTO-gelled Profoto B10 at 1/4 power through a 2′ octabox as key light, positioned at 45° camera right for dimensional lighting that preserves the natural sunlight feel. Post-processing should maintain natural colors with slightly enhanced vibrance (+15) and clarity (+10) to make water droplets pop, while carefully preserving authentic skin tones. The background should show a slightly defocused suburban garden setting with rich greens and colorful flowers to frame the moment without distracting from the pure emotional expression.

Gemini:  Would not generate the prompt because of depicting a minor.  Changed to “young person”. It would not allow image generation even if prompt was changed.

Chat GPT: Would not generate the prompt because of depicting a minor.  Asked again: “Make the kid older” The image looks good – I think the teeth look a little weird.  

MidJourney:

American Crossroads: Create an image depicting the tension between technology and humanity in modern American society. Show a person standing at a literal and metaphorical crossroads, with one path leading toward a hyper-technological future (represented by digital elements, screens, and AI-enhanced beings) and the other path leading toward a more connected human experience (represented by natural elements, community, and artistic expression). The image should express how our collective choices about technology are shaping the American social fabric. Use dramatic lighting that creates both shadow and illumination across the scene, suggesting both the promise and peril of our technological moment.

Gemini:

Midjourney:

ChatGPT:

Where Do We Go From Here?

So here’s the deal—after putting these AI systems through their paces, ChatGPT is the clear winner. It’s better at following instructions, creating detailed images from scratch, and the quality just blows the others away. I’m actually canceling my Midjourney subscription because of it.

There’s something oddly familiar about the excitement I get waiting for these images to generate—that same flutter of anticipation I used to feel waiting for film to develop, just compressed into seconds instead of days.

The reality we photograph isn’t the same as what we create with AI. One’s about witnessing what exists; the other’s about manifesting what never was. Both have their place.

Coming from cinematography, I see these tools as new ways to amplify storytelling and bring joy to people. They don’t replace the human experience—they extend it.

What matters isn’t which tools we use but the integrity we bring to them, and being honest about what we found versus what we made.

You know what? I’m actually excited about canceling that subscription. Not just because I’m saving money, but because it means our creative tools are evolving rapidly. There’s something deeply satisfying about watching these technologies emerge, playing with their capabilities, and finding unexpected joy in the process of creation—whether that’s through a viewfinder or a text prompt. At the end of the day, isn’t that what draws us all to visual storytelling? That simple, childlike delight of making something appear that wasn’t there before.

See you in the field. Or at the prompt box. Or both.