I personally think www.perplexity.ai provides the best prompt suggestions.
First Test
I asked for a prompt in English for “a bob-haired girl wearing a kendo uniform water skiing across a snowy winter landscape (side view),” and it suggested:
“A side view photo of a bob-haired girl wearing a traditional kendo uniform, water skiing across a snowy winter landscape. The scene captures the dynamic motion and unique contrast between her martial arts attire and the icy, snow-covered environment, with crisp winter light and a sense of energy and adventure.”
Here are the results:
1st Place: Dreamina
2nd Place: Gemini
3rd Place: ChatGPT
4th Place: Bing Image Creator (Copilot)
5th Place: Grok
Second Test
When I asked to write a prompt in English for “a cute cat holding grapes in its left hand and a smartphone in its right hand,” I got this response: “A photorealistic image of an adorable cat holding a bunch of grapes in its left paw and a smartphone in its right paw. The cat has a cute expression, fluffy fur, and is sitting or standing in a bright, minimal background with soft lighting. The scene captures a playful and charming moment, emphasizing the cat’s endearing pose and the contrast between the natural fruit and modern device.” I used this prompt.
1st Place: ChatGPT
2nd Place: Grok
3rd Place: Dreamina
4th Place: Gemini
5th Place: Image-FX
6th Place: Whisk
7th Place: Bing Image Create (Copilot)
Conclusion
1. Dreamina has great image quality and generates four images at once, which I like. (The downside is it might become paid soon…)
2. ChatGPT is hyped for its Ghibli-style images. It’s free but limited to a few images, slow, and generates only one image at a time, though the quality is good.
3. Gemini used to feel underwhelming, but it’s gotten much better recently. The downside is it only generates one image at a time.
4. Image-FX is actually close in ranking to 3rd and 5th. The issue is it often fails to generate images (which is why Image-FX and Whisk weren’t in Test 1). It ranks higher than Whisk because it generates four images at once.
5. Whisk has more fun features. The ability to upload an image and analyze its prompt is useful. However, no matter how good the prompt is, there’s a clear limit to the images it can produce.
6. Grok doesn’t seem bad, but it didn’t leave a strong impression on me personally. I’ll need to use it more to judge.
7. Bing Image Create (Copilot) struggles to produce realistic images, and I personally almost never use it.
