The new ChatGPT Images is here

Why This Update Matters

Before today, Google Gemini was dominating the AI image generation space with their Flash 2.5 and 3.0 models. But ChatGPT just leveled the playing field – and in some ways, surpassed it. This is a big deal because it means we now have real competition in the AI image space, which drives innovation and gives us better tools to work with.

I’ve been using Google Gemini’s image tools integrated directly into my WordPress media library through MinuteLaunch AI. Now I need to update those tools to include ChatGPT’s new image engine because the quality is just too good to ignore. This mandates that I update the MinuteLaunch image editor as well as the Media Library P2P.

The Tests: ChatGPT vs Google Gemini

Test 1: Outfit Swap

I started with a simple test – taking my profile picture and swapping my outfit. I uploaded the same circular profile image to both ChatGPT and Google Gemini with the same prompt: “Swap the person’s outfit.” The results were interesting: Gemini added some puka shells and kept me in the circle, but ChatGPT’s version looked incrementally better with more realistic lighting and texture. It’s not monumentally better, but it’s incrementally better.

Test 2: Miami Vice Transformation

This is where things got really fun. I took images of Fernando (my AI assistant based on Ricardo Montalbán) and myself and asked both platforms to create a Miami Vice 1980s scene. The prompt was: “Take these two guys and using their actual head style in Miami Vice 1980s unconstructed jackets on a Miami street corner in the middle of the evening with a Ferrari Testarossa, making them look like they’re ready to do police investigations.”

Interestingly, when I first tried this with Gemini, it just put Don Johnson in the image! Clearly it was confused by not having a recent copy of my face. I was using a circular profile image as a bit of a test to see if it could figure it out. After re-uploading my image, I tested both platforms. ChatGPT’s version had noticeably better photorealism.

Something amusing: Gemini flagged Fernando’s image as copyright infringement because of Ricardo Montalbán! Even though Fernando is my AI assistant and anybody who grew up in my era knows Ricardo Montalbán is a real (deceased) person.

Test 3: Car Mechanic Scene

This test really highlighted the differences. I asked both platforms to: “Turn this guy into a car mechanic and he’s leaning over the open hood of a V8 engine car, something like a Mercedes, checking the oil. And he’s wearing gloves and he is in the middle of a big automotive workshop. Give realistic workshop lighting and shadows so that you can see that he really appears to be under the car and the lights are affecting the shadows on his body differently.”

ChatGPT absolutely nailed this one. It put me in overalls with gloves, created incredibly realistic lighting effects, and even attempted to add my WP Launchify logo (though it turned it into “Lanch-fi” – it screwed up the letters but kept the logo design). The image looked like an actual photograph. If you didn’t know me, it would be really tough to tell this wasn’t real.

Gemini Flash 2.5’s version, while good, had that slightly painted, video game character quality that makes it feel less photorealistic. The softness in the way it renders makes it feel a little bit more like it’s painted. It definitely feels photorealistic to a degree, but this one from ChatGPT definitely feels more like an actual photograph.

Test 4: Glossy Photo Effects

One of ChatGPT’s unique features is the ability to create vintage photo effects. I asked it to: “Take the image of this reference person and make it into a photographic representation as if I was looking at a photograph that was printed on super glossy stock with a white border around it, a slight page curl and put the photo on a wood table top.”

The result was impressive – it created the shadow effect, added a page curl (a really neat effect), and maintained the photographic quality without changing my appearance. I can’t say that I see the glossy as much, but the page curl and shadow are there. This is the kind of effect you’d normally need Photoshop to create, but ChatGPT did it in seconds. It’s like looking at old photos from my wayward childhood in Law School – they show the edges of the photograph and so on. With digital, we’ve sort of lost that “oh, look, it’s on a Polaroid picture thrown on the table” feeling.

Adding Logos to Images

I tested ChatGPT’s ability to add logos to clothing by asking it to: “Take this image of the man and take this logo and put it in the upper right corner of his shirt as if it was embroidered in there.” I used my WP Launchify logo as the test.

The result was fascinating. ChatGPT did a reasonable job – not the best, but not the worst. It kept the logo design but changed the colors and scrambled the letters (turning “WP Launchify” into something like “Lanch-fi”). It’s not perfect for precise logo placement, but it shows the potential.

This gave me an interesting business idea for my Track Pilots community – I have a social network of about 12,700 track pilots with tons of images and videos that just sit there on my social network. I had an idea about how we can take those and combine them with rock and roll t-shirt themes to create custom merchandise using AI image generation. One of the ways we do that is of course having ChatGPT Images (or maybe Gemini) pulling off the actual compilation.

Inside MinuteLaunch’s WordPress Integration

While I was testing ChatGPT, I also showed how these tools are already integrated into MinuteLaunch AI’s WordPress media library. When you go to Media Library and click “Create AI Image,” you get access to Google Gemini 2.5 and 3.0 engines (and soon ChatGPT Images too).

The key advantage is that you can select any image from your media library as a starting point, use the built-in prompt creation helper for suggestions on lighting and composition, and generate images that are automatically saved to your WordPress media library. No copying and pasting between platforms – everything happens right where you need it. It’s nice because it’s already in my media library ready to use.

I use this tool almost every day to generate thumbnails for my YouTube videos. The prompt helper gives suggestions like: “Create a high resolution photograph of a young African-American male in a bustling automotive workshop with various tools scattered around. He is dressed in a fitted gray mechanic shirt, blue jeans, engaged in checking the oil.” These little differences really do help. Obviously if I want to do the one of me, I’ll just change the reference image.

The Photorealism Difference

When I compared the final car mechanic images side-by-side, the difference was clear. ChatGPT’s version looked like it could be a real photograph – if you didn’t know me, it would be really tough to tell it wasn’t real. The lighting, shadows, and texture all felt authentic. Really interesting lighting effect.

Gemini Flash 2.5’s version, while still impressive, had that slightly softer, painted quality that makes it feel more like a video game character. It’s not bad – it’s actually quite good – but for professional content where photorealism matters, ChatGPT has taken a significant leap forward.

Holiday Cards and Creative Applications

Just for fun, I also created some holiday greeting cards using ChatGPT Images. I did my holiday greeting cards early. The results were so good that I wouldn’t be embarrassed to actually send them out as my Hanukkah holiday cards this year. It’s like my mom took me to Walmart or Kmart for those classic family photo sessions – except it’s all AI-generated. These are really quite awesome to be honest with you. I would not be embarrassed to send this out for my holiday cards this year. Put them in an order and I’m ready to go.

I even created a bubble head version of myself, which was amusing if not entirely practical. But it shows the range of creative possibilities these tools unlock. There’s lots of capability here.

What This Means for Your Business

We’re in an arms race right now with AI image generation. Every few weeks, these platforms are releasing major improvements that were impossible just months ago. What I’m doing is incorporating these tools into systems you own and control – either through MinuteLaunch AI or FluentMCP.

The kind of output and outcome that required a huge budget in the past is now possible with a minimal budget – just a couple hundred bucks a month. You can generate professional images, create custom graphics, transform photos, and produce content that would have required a professional designer and photographer.

I’m working on two, three, four, five different things every day that are designed to incorporate these capabilities into the box you own and control. That’s the key – you’re not dependent on some external platform that could change their pricing or terms tomorrow. You own the WordPress site, you control the tools, and you can use them however you need. In doing so you can have the kind of output and outcome that quite honestly was just not possible even with a huge budget in the past. But now it’s possible with nearly a minimal budget, right? A couple hundred bucks a month.