For the longest time, image generation was one of the few areas where Gemini had an edge over ChatGPT. Google’s Imagen model generated more realistic images and followed prompts better, whereas OpenAI’s DALL·E often produced images that looked more AI-ish and cartoonish. But that’s about to change.
Now, both Google and OpenAI are rolling out native image generators powered by their multimodal AI models. While Gemini’s version is still tucked inside AI Studio — Google’s beta testing platform — OpenAI has released its native image generator directly inside ChatGPT.
Here’s the twist. Gemini’s version is available to everyone for free, while OpenAI delayed the rollout for free users due to excessive demand. I tried both. And let’s just say the tables may be turning.
Table of Contents
1. Changing the Style of an Image
One of the biggest perks of native image generation is that it’s not just about creating images from scratch—it’s also great at editing or transforming existing ones. After ChatGPT rolled out this feature, X was flooded with anime-style images inspired by Studio Ghibli. So I decided to try the same. I uploaded a photo of a guy into both ChatGPT and Gemini, asking them to convert it into anime Ghibli style. Here’s what I got:
Let’s just say this clearly: Gemini struggles to change the style completely. It mostly sticks to the original image, often just bumping up the brightness regardless of what you ask. In contrast, ChatGPT transforms the image—anime style, pixel art, lego characters, Simpsons, 3D pixar style animation, you name it. While it changes a few details, especially faces, the overall result is miles ahead of other models.
Verdict: Only ChatGPT can fully transform an image into a different style. Gemini’s native image generator currently falls short
2. Editing Small Changes on Top of Image
This is where Gemini shines. I uploaded the same image to both models and asked them to add eyeglasses. Here’s what happened:
Both got the job done—but in different ways. ChatGPT tends to redraw the entire image, sometimes even altering the person’s face. In contrast, Gemini simply adds the glasses without touching anything else. That’s because ChatGPT still generates a new image based on the original, whereas Gemini can make edits on top of the original. Similarly, you can also remove objects, etc. in Gemini. People are already using this feature to remove watermarks and various crazy stuff.
Verdict: Gemini wins if you want clean edits without changing the original image.
3. Generating a Realistic-Looking Image From Scratch
This used to be a weak spot for ChatGPT—but not anymore. I asked both models to generate a close-up of an old man in his 70s, wearing a soft wool cardigan over a white shirt. Here’s what they came up with:
Both nailed the prompt and are pretty much on par. ChatGPT’s image is slightly more polished, while Gemini’s version feels more realistic, capturing subtle camera imperfections and a natural look. ChatGPT, by contrast, leans toward perfection—almost too perfect at times. However, choosing one over the other can be a personal preference.
Verdict: Both models can generate realistic images with ease.
4. Blending Two Images Into One
Both ChatGPT and Gemini now let you upload multiple reference photos to generate a new image. In this test, I uploaded a photo of a man and a separate image of another man wearing a green shirt. Finally, asked both models to generate a picture of the first man wearing the green shirt. Here’s what I got:
ChatGPT delivers consistently good results. Gemini, however, sometimes skips the head or outputs a low-quality image—but those glitches usually go away with a retry or two. On the flip side, Gemini nails the shirt color more accurately, while ChatGPT introduces a slight variation.
Verdict: ChatGPT wins for consistency and overall quality. But Gemini’s not far behind
5. Generate an image with a Different Point of View
Both ChatGPT and Gemini can also generate images from a different point of view. For this test, I uploaded a photo of a train’s interior and asked both models to recreate the scene from the opposite side.
Both delivered decent results but struggled with object placement, especially in complex images with many elements. That said, if following such details isn’t important and you’re just looking for a fresh perspective of a subject—like a car or a building—both models handle it well.
Verdict: Both models did a decent job but messed with the placement of objects.
6. Generating a Birthday Card
These new models are also said to be much better at generating images with text — a task older models often struggled with. So, I decided to test that by generating a birthday card with specific text.
Surprisingly, both models nailed it. They followed the prompt exactly — using cursive “Happy Birthday” text and surrounding it with floral designs, just as asked. Which one looks better is really a matter of personal taste. We also tried generating menus, placards, infographics, and other text-heavy images — and both models handled them quite well.
Verdict: Both Gemini and ChatGPT can now generate accurate, readable text within images and follow prompts closely.
ChatGPT vs Gemini Native Image Generator
Across all our tests, a few patterns stood out. ChatGPT consistently delivers higher-quality images — sharper details, better composition, and fewer weird artifacts. With Gemini, you often have to regenerate a few times before landing on something good. That said, Gemini is noticeably faster. It can produce an image in around 10 seconds, while ChatGPT can sometimes take a minute or more, even for simple prompts.
When it comes to generating realistic visuals, handling text in images, or switching perspectives, both models perform similarly. But the real difference shows up during edits. ChatGPT is much better at transforming the overall style of an image, while Gemini shines when you want to add or remove specific objects without changing anything else.
Overall: ChatGPT offers more consistent quality and a better all-around experience — if you don’t mind the wait.
