I attempted ChatGPT Images 2.0: A enjoyable, large leap – and surprisingly helpful for actual work

Follow ZDNET: Add us as a preferred source on Google.

---

ZDNET’s key takeaways

• Images 2.0 delivers correct textual content and usable graphics.
• It can match model kinds, together with ZDNET visuals.
• Errors nonetheless slip in, requiring human evaluation.

---

Earlier this week, OpenAI unveiled ChatGPT Images 2.0, its new picture era engine. Key to this launch is a bounce in performance from creating “decorations” (OpenAI’s time period) to full-page graphics, together with detailed textual content.

I had early entry to a pre-release model. It labored fairly properly, however saved messing up on the ZDNET emblem. Now that the product has been formally launched, I’m giving it an in-depth check throughout a variety of challenges.

Images 2.0 is offered to all ChatGPT tiers, however the extra succesful language options are solely obtainable to paying tiers that may use the Thinking mannequin. I’m working all these exams utilizing a ChatGPT Plus account with Thinking turned on.

Also: I put GPT-5.5 by way of a 10-round check: It scored 93/100, dropping factors just for exuberance

Let’s get began with the ZDNET branding workouts. Rather than simply importing ZDNET pages and having it discover the brand on the web page, I created a standalone picture of the ZDNET emblem and uploaded that with every immediate. That appeared to assist tremendously.

[One quick note: ZDNET doesn’t permit OpenAI to scrape its pages. Ziff Davis, ZDNET’s parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems. So I used a Chrome extension to capture full-screen screenshots of the articles I wanted to test with Images 2.0. That’s how ChatGPT was able to read them.]

Can Images 2.0 protect the ZDNET emblem?

My start line was the article I beforehand wrote about Images 2.0. I fed ChatGPT this immediate: “Create a detailed and vivid infographic of this article using the ZDNET brand style and the attached ZDNET logo.”

Not solely is the brand appropriate, however the coloring is ideal for ZDNET. But the place the picture actually shines is its use of textual content. All the textual content is appropriate, even the tiny textual content on an angle within the picture.

Can it produce styled sketchnotes?

Next, I made a decision to revisit the sketchnotes problem I gave to Google’s Nano Banana a couple of months in the past. The project at the moment was to create a sketchnotes model of the US Bill of Rights. Nano Banana did an excellent job with the pictures, however I needed to attempt time and again (and over) to persuade it to get the wording proper. Read the article to see the hoops I needed to bounce by way of.

Also: I used Nano Banana 2 to make good sketchnotes: 5 classes realized

For ChatGPT Images 2.0, I upped the stakes barely. I wished sketchnotes, however I wished them in ZDNET’s branding type. I’m taking part in up the branding type all through this text as a result of that is a method ChatGPT Images 2.0 may present actual worth to customers.

Here’s the primary immediate: “Make me a sketchnote of the US Bill of Rights. Use the ZDNET logo style and make the sketchnotes in the ZDNET style.” That’s the picture on the left. Here’s the second immediate: “Include the ZDNET logo and add more neon-style colors, perhaps on a black background.” That’s the picture on the precise.

First, discover that the textual content is appropriate. There are not any duplicates. Nothing is lacking. Already, that is head and shoulders above Nano Banana’s efficiency. Both variations match with ZDNET’s type. The solely factor I’m not thrilled with is that the ZDNET emblem seems jammed in on the second picture. Even so, the brand is appropriate, and I may in all probability do a couple of extra immediate passes to get it positioned higher.

Wacky enjoyable with an infographic

But now we come to the unforced error my testing set revealed. I requested Images 2.0 to transform my AI web site builder shootout article to an infographic. It produced a reasonably usable, if considerably busy, infographic. It even went to the web and added data I did not have within the article, like base pricing.

But there are 4 clear errors:

1. The header highlights “here are 9 of the best AI website builders.” It even makes the “9” stand out. Except that solely 5 web site builders have been reviewed. Lower within the infographic, it reveals the 5 I do evaluation. Oops.
2. The companies I reviewed have been Hostinger, GoDaddy, Wix, 10Web, and Squarespace. ChatGPT determined, for some motive, to switch 10Web with Durable (a competitor to 10Web). I did not evaluation Durable. I did not even point out Durable. Wacky.
3. The AI produced a abstract desk for the companies, itemizing star scores for ease of use, design flexibility, and AI options. But I did not present star scores for these classes. The AI was overly beneficiant towards some distributors, in a means that was immediately opposite to the evaluation textual content itself. Odd.
4. Finally, and it is a nit, however nonetheless. Way down on the backside, the place the AI appropriately reproduced the ZDNET emblem, there is a drooping line simply above it. Why?

Also: The finest AI picture turbines: There’s just one clear winner now

To be honest, these are all errors an in-house human graphic designer would possibly produce in a primary draft. In my years as a founder and a product supervisor, I’ve definitely seen extra egregious graphics errors come again from my designers on their first drafts.

When I re-prompted Images 2.0 with corrections (apart from the star scores, which I did not appropriate within the second picture), it did appropriately modify the infographic with extra acceptable data.

ChatGPT Images has come a great distance

This Images 2.0 launch is a large enchancment over earlier variations. The ChatGPT Images model I checked out final 12 months was spectacular, particularly for recontextualizing photos.

Also: I bought an early have a look at ChatGPT Images 2.0, and it is spectacular – with one exception

This new model, which might interpret precise content material after which create photos, is a large leap over earlier builds. More to the purpose, it could possibly ship very tangible enterprise worth, which makes it value rather a lot not just for enjoyable footage however for actual work.

Stay tuned, as a result of I’ll be how this construct compares with Google Gemini’s Nano Banana. I’ll be pushing it even additional to see what different work-related duties it could possibly assist with, significantly in terms of consumer interface design.

How snug are you counting on AI-generated visuals, figuring out that the mannequin can introduce delicate factual errors? Let us know within the feedback beneath.

---

You can observe my day-to-day undertaking updates on social media. Be certain to subscribe to my weekly update newsletter, and observe me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/DavidGewirtz, on Instagram at Instagram.com/DavidGewirtz, on Bluesky at @DavidGewirtz.com, and on YouTube at YouTube.com/DavidGewirtzTV.