How experts use Images 2.0 in their work
Learn from a mathematician, a high school teacher, a demolition professional, and more
We recently launched ChatGPT Images 2.0, a state-of-the-art model that can take on complex visual tasks and produce immediately usable images.
Images 2.0 is our first image model with thinking capabilities. When a thinking or pro model is selected in ChatGPT, Images 2.0 can search the web for real-time information, create multiple distinct images from one prompt, and double-check its own outputs.
We’re sharing tips and five examples of how professionals are experimenting with this new model in their fields.
1. Add technical annotations to a photo
“This is a wiring diagram for a control box I built to manage temperature in my vinyl greenhouses. The Japanese came out perfectly. I was genuinely surprised.”
— Hiroki Tomiyasu, farmer (northern Japan)
His prompt*:
This is the inside of the control panel for a roll-up machine for a vinyl greenhouse that I’m building. I’d like you to create an image that explains what each part is, what machine it belongs to, and how the whole system works. The component visible in the upper left is the terminal block, the one in the center is the motor driver, the upper right is the ESP32, and there are switches in the lower right and lower left corners.
[Wires.jpg]
*Translated from Japanese
2. Visualize complex processes
“I asked ChatGPT to make a step-by-step guide to starting a small construction business based on my experiences, and it’s exactly what I wished I had when I started. Twenty books about business and companies and construction, and they didn’t clearly explain how to do it from a to z.”
— James Costello, demolition professional (U.S.)
His prompt:
Generate an image explaining how to start a small construction business, step by step. Use my attached notes as a reference.
[Notes.pdf]
3. Make a realistic photo of a scientific concept
“I like how well the physics is represented in this model. The standing Faraday waves that such fluids generate are incredibly hard to visualize, and this is a realistic image of the actual situation.”
— Bartosz Naskręcki, PhD, mathematician and professor (Poland)
His prompt:
Make a realistic photo of the non-Newtonian liquid under a heavy acoustic signal.
4. Create shareable artifacts for your students
“One of the biggest improvements I’ve noticed is how much better the model has become with math. I usually look at how it handles math as a litmus test, because math really shows whether the model can be both creative and accurate. It’s not enough for something to just look good. The math has to be clear, correct, and actually useful. I am in the process of thinking about how to incorporate these playing cards into our end-of-course review.”
— Racquel Gibson, high school algebra teacher (U.S.)
Her prompt:
Create a set of 4 collectible trading cards featuring the characteristics of linear, absolute value, exponential, and quadratic functions in a pixel art style. Generate one image per card.
5. Create a design system based on an example
“Very powerful use case on my end to create an in-depth design system one-pager as I’m working on our new website redesign for Protopica. Some very minor glitches on icons and typos, but besides that it’s 99% accurate.”
— Manuel Sainsily and Will Selviz, creative directors (Canada)
Their prompt:
Based on this website, create a one-pager covering the entirety of the design system: typography, colors, visual DNA, buttons, icons, spacing, scale, material, motion, etc.
[index.html]
Pro tips
Here are a few tips from users about how to best work with Images 2.0:
Tip 1: Give it a reference image to model the style you are looking for.
Images 2.0 is better able to capture the defining characteristics of photos, and as a result, can produce outputs that more faithfully reflect the style requested, rather than approximating it.
Tip 2: Play with different aspect ratios and generate multiple images at once.
Images 2.0 supports aspect ratios as wide as 3:1 and as tall as 1:3 and you can ask for a coherent set of up to eight outputs in one go.
Tip 3: Make images in many languages.
Images 2.0 has stronger multilingual understanding and non-Latin text rendering, particularly in Japanese, Korean, Chinese, Hindi, and Bengali.
Translated excerpt from Korean:
This latest ChatGPT Image 2.0 is truly impressive. You’ll understand once you try it. Not only does it execute prompts accurately, but even more surprising is how well it handles cultural context and text generation. For example, I asked for the background of the first image to be COEX, and the third image to be Rotterdam Centraal Station in the Netherlands. Incredible — suddenly it recreated Rotterdam Centraal so accurately, even down to the clock, which made me unexpectedly nostalgic.
Tip 4: Try putting the image in a new chat if you get stuck in the editing process.
For more stories and use cases from experts using OpenAI’s technology, subscribe.
Are you a Pro subscriber using ChatGPT and Codex to help you level up your work? We’d love to meet you.