OpenAI’s new GPT Image 1.5 makes ChatGPT’s image generation noticeably faster and, in some cases, noticeably better. It improves how consistently the system handles details like lighting and facial features across edits, and it does a better job rendering text and following instructions. GPT Image 1.5 is now available to all ChatGPT users, both paid and free.
The release comes as OpenAI and Google compete for dominance in AI, particularly in image generation. Google recently released Gemini 3 Pro with Nano Banana Pro, which reportedly triggered a “code red” response inside OpenAI. GPT Image 1.5 is OpenAI’s answer to that new competition.
What’s Actually Better in GPT Image 1.5
In my testing, GPT Image 1.5 handled complex prompts better than the previous version and generated more realistic-looking images overall. It also performs well at text rendering. I asked it to generate an image of an iMessage conversation on an iPhone, and it got the UI elements right along with nearly perfect text. It is also more reliable at following instructions like “remove this person” or “change the background to a beach.”
But it is not perfect. When I asked ChatGPT to create an image in a 16:9 aspect ratio, it produced something only slightly wider than a square image, and nowhere near the format I requested. Google’s Gemini nailed the 16:9 ratio with the same prompt. That kind of inconsistency is frustrating when you need specific dimensions or layouts.
For example, I gave both Gemini Nano Banana Pro and GPT Image 1.5 the same detailed prompt, and these are the results.
AI-generated images still have a fundamental problem. They often look obviously fake. GPT Image 1.5 does better than earlier versions at getting small details right, and that helps with realism. But you can still usually tell when an image was generated by AI, especially when people are involved.
We tested ChatGPT’s previous image tools against Google’s Gemini in October and found that ChatGPT was better for complex editing tasks like filling in missing details, while Gemini was faster and better at maintaining character consistency. With GPT Image 1.5, OpenAI has closed the speed gap somewhat, although Gemini still feels faster overall. GPT Image 1.5’s improvements make it more useful, especially if you iterate a lot on images. But Gemini still follows some instructions more precisely.
If you already rely on ChatGPT for other work, the improved image generation is worth trying. Just do not expect it to replace other tools yet.
[Image credit: Palash Volviokar/Techlicious via Nano Banana Pro and ChatGPT]
