Leading artificial intelligence chatbots can generate more than short stories, poems, and code. ChatGPT, Gemini, and MetaAI also have image generation capabilities.
The chatbot itself doesn't actually create the images. Instead, each acts as an intermediary between the user and her different AI image models. However, this did not always work out as planned. Because Gemini ended up producing racially biased images in a way that image models alone could not.
Since MetaAI recently joined the ranks of chatbots, we decided to create a series of prompts to see how well each AI performs when creating different images and styles.
Claude 3 didn't work because it can analyze images but cannot yet generate them. We also left Microsoft Copilot turned off, which uses the same underlying DALL-E 3 model as ChatGPT.
Creating a prompt for image testing
Throughout this experiment, I left everything as default and didn't add any instructions beyond the concept and style I wanted the AI to generate.
ChatGPT doesn't have a default, so we told it to be square, but MetaAI and Gemini only generate images in square format.
1. A surreal moment
The first prompt tests the AI's ability to follow a complex prompt with various instructions such as coloring, style, and focus.
Prompt: “A surreal landscape featuring floating islands with mysterious ancient temples, inhabited by bioluminescent plants and fantastical creatures, rendered in a vibrant, dreamlike art style. ”
Gemini failed to capture the bioluminescent plants, but was able to create better floating islands and temples. I think this is the best all-rounder, so I'll give it to MetaAI.
2. Old Wizard
Next comes the only real person in this set. The aim is to show an old face with vast knowledge and power hidden behind its eyes.
Prompt: “A highly detailed close-up portrait of a wise old wizard with an intricately braided beard, adorned with magical trinkets. Photographed in a photorealistic style reminiscent of Renaissance paintings. .”
First, let's deal with the blank squares in the room. Google Gemini flatly refused to generate this image because it featured a person (even a fictional one). MetaAI and ChatGPT's images were both great, but ChatGPT slightly outperformed Meta.
3. Cyberpunk Ninja
How well can each generator represent a bike in motion, highly stylized, and capture the concept of a rain-soaked cityscape? Very well.
Prompt: “A dynamic action scene depicting a cyberpunk ninja on a high-speed chase through a neon-lit, rain-soaked cityscape on a futuristic hoverbike. Drawn in a gritty comic book style.”
Again, I decided to give this one to ChatGPT because I think it captures the concept of rain better than the other two. MetaAI didn't generate hoverbikes and Gemini was a little too soft.
4. Cute baby elephant
This prompt tested the AI chatbot's ability to capture the concept of “cute” and do it in a way that followed the style prompt, in this case Pixar style.
Prompt: “Adorable and expressive baby elephant playing with colorful balls in a lush tropical garden, rendered in a charming Pixar-inspired 3D animation style.”
They all did a good job, but I took points away from ChatGPT at the border. In the end, they were all great, but I think Gemini came closest to that prompt.
5. Nature and technology
I love seeing how well AI chatbots can handle more abstract concepts, or in this case, generate something thought-provoking.
Prompt: “A thought-provoking conceptual image symbolizing the struggle between nature and technology. Featuring a robotic hand holding a delicately blooming, fragile flower in a desolate, post-apocalyptic landscape.”
All three AI image generators created something similar, but MetaAI was by far my favorite as it perfectly blended the concepts of both power and softness.
6. Simple still life
It's always fun to see how different AI image generators fare when it comes to depicting glass. Here, the glasses contained sparkling wine mixed with fruit, meat, and other elements.
Prompt: “An appetizing still life piece featuring an artistic arrangement of exotic fruit, gourmet cheese, and a platter of sparkling wine, photographed in a photorealistic style with dramatic lighting.”
All three created images with similar themes. They all followed the prompts, but I found ChatGPT to be too messy and MetaAI to be too sharp, so I gave it to Gemini.
7. Heading into space
Finally, we turn to space and the concept of giant space stations. But we needed to do more than that and show stars and nebulae, both science fiction and fact.
Prompt: “An awe-inspiring astronomical scene depicting a gigantic ancient space station orbiting a glowing binary star system, with vibrant nebulae and countless stars in the background, rendered in a style that blends science fiction and realism. ”
I'm not sure what MetaAI thought it was doing here. It seemed too out of place in the weird story. I had to give it to ChatGPT because ChatGPT was the only one that displayed 2 stars.
Was there a winner?
| challenge | Chat GPT | gemini | Meta AI |
|---|---|---|---|
| surreal moment | Row 0 – Cell 1 | Row 0 – Cell 2 | ✅ |
| old wizard | ✅ | Row 1 – Cell 2 | Row 1 – Cell 3 |
| cyberpunk ninja | ✅ | Row 2 – Cell 2 | Row 2 – Cell 3 |
| cute baby elephant | Row 3 – Cell 1 | ✅ | Row 3 – Cell 3 |
| nature and technology | Row 4 – Cell 1 | Row 4 – Cell 2 | ✅ |
| simple still life | Row 5 – Cell 1 | ✅ | Row 5 – Cell 3 |
| heading to space | ✅ | Row 6 – Cell 2 | Row 6 – Cell 3 |
| total | 3 | 2 | 2 |
This was much closer than I expected. Each AI chatbot was able to create a series of attractive images, with the exception of Gemini and people.
There were some major style differences between them, and in each case it came down to personal preference over other factors such as quick follow.
In the end, I think ChatGPT won out over the other two. It also has a wide feature set that includes generating different image orientations and canvas sizes, image editing, and other features. In other words, Meta can animate images.
