I tried ChatGPT Images 2.0: A fun, huge leap - and surprisingly useful for real work

David Gewirtz / Elyse Betters Picaro / ZDNET

Follow ZDNET: Add america arsenic a preferred source connected Google.

ZDNET's cardinal takeaways

• Images 2.0 delivers close substance and usable graphics.
• It tin lucifer marque styles, including ZDNET visuals.
• Errors inactive gaffe in, requiring quality review.

Earlier this week, OpenAI unveiled ChatGPT Images 2.0, its caller representation procreation engine. Key to this merchandise is simply a leap successful functionality from creating "decorations" (OpenAI's term) to full-page graphics, including elaborate text.

I had early entree to a pre-release version. It worked rather well, but kept messing up connected the ZDNET logo. Now that the merchandise has been officially released, I'm giving it an in-depth trial crossed a wide scope of challenges.

Images 2.0 is disposable to each ChatGPT tiers, but the much susceptible connection features are lone disposable to paying tiers that tin usage the Thinking model. I'm moving each these tests utilizing a ChatGPT Plus relationship with Thinking turned on.

Also: I enactment GPT-5.5 done a 10-round test: It scored 93/100, losing points lone for exuberance

Let's get started with the ZDNET branding exercises. Rather than conscionable uploading ZDNET pages and having it find the logo connected the page, I created a standalone representation of the ZDNET logo and uploaded that with each prompt. That seemed to assistance tremendously.

[One speedy note: ZDNET doesn't licence OpenAI to scrape its pages. Ziff Davis, ZDNET's genitor company, filed an April 2025 suit against OpenAI, alleging it infringed Ziff Davis copyrights successful grooming and operating its AI systems. So I utilized a Chrome hold to seizure full-screen screenshots of the articles I wanted to trial with Images 2.0. That's however ChatGPT was capable to work them.]

Can Images 2.0 sphere the ZDNET logo?

My starting constituent was the nonfiction I antecedently wrote astir Images 2.0. I fed ChatGPT this prompt: "Create a elaborate and vivid infographic of this nonfiction utilizing the ZDNET marque benignant and the attached ZDNET logo."

David Gewirtz via ChatGPT Images/ZDNET

Not lone is the logo correct, but the coloring is cleanable for ZDNET. But wherever the representation truly shines is its usage of text. All the substance is correct, adjacent the tiny substance connected an space successful the image.

Can it nutrient styled sketchnotes?

Next, I decided to revisit the sketchnotes challenge I gave to Google's Nano Banana a fewer months ago. The duty astatine that clip was to make a sketchnotes mentation of the US Bill of Rights. Nano Banana did a large occupation with the images, but I had to effort implicit and implicit (and over) to person it to get the wording right. Read the nonfiction to spot the hoops I had to leap through.

Also: I utilized Nano Banana 2 to marque cleanable sketchnotes: 5 lessons learned

For ChatGPT Images 2.0, I upped the stakes slightly. I wanted sketchnotes, but I wanted them successful ZDNET's branding style. I'm playing up the branding benignant passim this nonfiction due to the fact that that's 1 mode ChatGPT Images 2.0 could supply existent worth to users.

Here's the archetypal prompt: "Make maine a sketchnote of the US Bill of Rights. Use the ZDNET logo benignant and marque the sketchnotes successful the ZDNET style." That's the representation connected the left. Here's the 2nd prompt: "Include the ZDNET logo and adhd much neon-style colors, possibly connected a achromatic background." That's the representation connected the right.

David Gewirtz via ChatGPT Images/ZDNET

First, announcement that the substance is correct. There are nary duplicates. Nothing is missing. Already, this is caput and shoulders supra Nano Banana's performance. Both versions acceptable with ZDNET's style. The lone happening I'm not thrilled with is that the ZDNET logo looks jammed successful connected the 2nd image. Even so, the logo is correct, and I could astir apt bash a fewer much punctual passes to get it placed better.

Wacky amusive with an infographic

But present we travel to the unforced mistake my investigating acceptable revealed. I asked Images 2.0 to person my AI website builder shootout article to an infographic. It produced a reasonably usable, if somewhat busy, infographic. It adjacent went to the net and added accusation I didn't person successful the article, similar basal pricing.

David Gewirtz via ChatGPT Images/ZDNET

But determination are 4 wide errors:

1. The header highlights "here are 9 of the champion AI website builders." It adjacent makes the "9" basal out. Except that lone 5 website builders were reviewed. Lower successful the infographic, it shows the 5 I bash review. Oops.
2. The services I reviewed were Hostinger, GoDaddy, Wix, 10Web, and Squarespace. ChatGPT decided, for immoderate reason, to regenerate 10Web with Durable (a rival to 10Web). I didn't reappraisal Durable. I didn't adjacent notation Durable. Wacky.
3. The AI produced a summary array for the services, listing prima ratings for easiness of use, plan flexibility, and AI features. But I didn't supply prima ratings for these categories. The AI was overly generous toward immoderate vendors, successful a mode that was straight contrary to the reappraisal substance itself. Odd.
4. Finally, and this is simply a nit, but still. Way down astatine the bottom, wherever the AI correctly reproduced the ZDNET logo, there's a drooping enactment conscionable supra it. Why?

Also: The champion AI representation generators: There's lone 1 wide victor now

To beryllium fair, these are each errors an in-house quality graphic decorator mightiness nutrient successful a archetypal draft. In my years arsenic a laminitis and a merchandise manager, I've surely seen much egregious graphics errors travel backmost from my designers connected their archetypal drafts.

When I re-prompted Images 2.0 with corrections (except for the prima ratings, which I didn't close successful the 2nd image), it did correctly modify the infographic with much due information.

ChatGPT Images has travel a agelong mode

This Images 2.0 merchandise is simply a immense betterment implicit erstwhile versions. The ChatGPT Images mentation I looked astatine past year was impressive, particularly for recontextualizing images.

Also: I got an aboriginal look astatine ChatGPT Images 2.0, and it's awesome - with 1 exception

This caller version, which tin construe existent contented and past make images, is simply a immense leap implicit erstwhile builds. More to the point, it tin present precise tangible concern value, which makes it worthy a batch not lone for amusive pictures but for existent work.

Stay tuned, due to the fact that I'll beryllium looking astatine however this physique compares with Google Gemini's Nano Banana. I'll beryllium pushing it adjacent further to spot what different work-related tasks it tin assistance with, peculiarly erstwhile it comes to idiosyncratic interface design.

How comfy are you relying connected AI-generated visuals, knowing that the exemplary tin present subtle factual errors? Let america cognize successful the comments below.

You tin travel my day-to-day task updates connected societal media. Be definite to subscribe to my play update newsletter, and travel maine connected Twitter/X astatine @DavidGewirtz, connected Facebook astatine Facebook.com/DavidGewirtz, connected Instagram astatine Instagram.com/DavidGewirtz, connected Bluesky astatine @DavidGewirtz.com, and connected YouTube astatine YouTube.com/DavidGewirtzTV.