OpenAI released GPT-Image-2 yesterday and it is impressive. Really impressive and it now supports 4k which is awesome. The state of image generation in general has been getting better and better and I now use models to do some cool things on a regular basis. Here are some examples that are super powerful - all generated with GPT-Image-2:

Personalized Art

I love using AI to create art for the right thing at the right time. Its cool because I can create things I like and tailor them endlessly. It also allows me to be creative to spur other creative ideas.

Prompt: Set in dusk, with a setting sun, there is a road that disappears over the horizon with a single car heading towards the viewer. The sides of the road are overgrown and feel abandoned and littered with some futuristic elements. There is a sign that says New York City, 300 miles and it is worn.

Prompt: Set in dusk, with a setting sun, there is a road that disappears over the horizon with a single car heading towards the viewer. The sides of the road are overgrown and feel abandoned and littered with some futuristic elements. There is a sign that says New York City, 300 miles and it is worn.

Infographics

Gpt-image-2’s ability to create infographics is awesome. Text is clear, crisp and even with simple prompts you can turn a topic into a nice visual representation. This will be hugely beneficial for improved learning and accelerated understanding. Here is one I generated that helps explain the quantum superposition. This is insanely useful.

Prompt: I want an infographic that explains what quantum superposition means and why it is important

Prompt: I want an infographic that explains what quantum superposition means and why it is important

I even had it build an infographic on how headshots can be created using AI.

Prompt: Create an infographic that explains how AI headshot generators work - the ones that you give it some photos of yourself and it can generate a professional style headshot

Prompt: Create an infographic that explains how AI headshot generators work - the ones that you give it some photos of yourself and it can generate a professional style headshot

Headshots

GPT-Image-2 does great at creating professional headshots too with a few sample images. I can provide a sample image and some direction and it will generate a professional looking headshot rapidly.

Prompt: Generate a professional headshot from this starting image with a professional background

Prompt: Generate a professional headshot from this starting image with a professional background

Sankey Generation

GPT-Image-2 does a great job of creating sankey diagrams. I have often seen those around company financial statements and they have been a great visual to breakdown a company’s income. I tried it with a simple prompt and pointed to raw sample dataset. It did a great job at least directionally. (though I have not validated all the numbers here, so it might need some refinement/correction).

Prompt: create a sankey diagram from this dataset: https://raw.githubusercontent.com/micahstubbs/sankey-datasets/refs/heads/master/oakland-budget/graph.json

Prompt: create a sankey diagram from this dataset: https://raw.githubusercontent.com/micahstubbs/sankey-datasets/refs/heads/master/oakland-budget/graph.json

Just for fun

It can even generate comic strips for you in the style of your choosing. Here I took a fun interaction I had with my daughter and turned it into a memory I could frame - kid’s sometimes say the best things.

Prompt: I want to create a 3 pane comic panel in the style of calvin & hobbes based on a fun story between my and my daughter. We were walking in a quaint town and she said Hey Dad, Did you know that Jesus died playing lacrosse…a boy in my class told me that

Prompt: I want to create a 3 pane comic panel in the style of calvin & hobbes based on a fun story between my and my daughter. We were walking in a quaint town and she said Hey Dad, Did you know that Jesus died playing lacrosse…a boy in my class told me that