A recent Reddit thread has sparked a heated debate about the image quality of Google Gemini's Nano Banana 2 model. Many users, including Pro subscribers, claim that the image generator has degraded over time. One user stated that almost every image appears flat or cartoonish, even with highly detailed prompts. These complaints, which have been circulating since May, suggest a widespread perception of regression, but is it really the case?
Reddit complaints and possible causes
Nano Banana 2 has not received any official updates since February, while the latest change to Gemini dates back to July with the introduction of Gemini 3.6 Flash. However, Google has never announced changes to the image generator. Despite this, users report noticeable differences. A plausible explanation is that even if the generation model remained the same, the text model interpreting prompts may have changed, thus altering the final output. This hypothesis is supported by the fact that complaints started before the latest Gemini update.
Sponsored Protocol
A journalist's comparative test with realistic prompts
To verify the claims, a journalist conducted a personal test comparing Nano Banana 2 and ChatGPT on three different prompts. In the first test, a realistic dragon flying overhead, ChatGPT produced an image that looked like a real photo taken from below, while Gemini generated a dragon that was more posed and with an overall quality that hinted at an "AI-generated" look. In the second test, a couple in their 50s in a café, ChatGPT delivered a more natural result, while Gemini showed misplaced details, such as wrong reflections and an uncanny valley effect on faces. In the third test, a full English breakfast, both performed well, but Gemini was let down by unreadable text on the menu.
Sponsored Protocol
Image quality and the perception of decline
The journalist concludes that the "flat and cartoonish" criticism toward Nano Banana 2 is not entirely justified. Rather than a worsening of Google, it is more likely that ChatGPT has improved significantly over the past few months. OpenAI has made huge strides in image generation, and if ChatGPT has advanced while Nano Banana 2 has remained essentially unchanged, the subjective impression could easily be that Google has gotten worse, when in reality it has just been overtaken. Although ChatGPT is slower, its images appear more realistic and less artificial, making it the preferred choice for those seeking high-quality results.
Sponsored Protocol
This evolution raises questions about the future of AI image generation. While Google integrates Nano Banana 2 into products like Google Earth, competition pushes OpenAI to innovate faster. For users, the choice between the two tools depends on priorities, but the test suggests that for photorealism, ChatGPT is currently superior. Of course, perception is subjective, and others may prefer Gemini's style. It remains to be seen whether Google will respond with a substantial update to Nano Banana 2 or if OpenAI's advantage becomes an unbridgeable gap.
