ChatGPT image generation has gotten a lot better with GPT-4o.
If you still struggle to get the types of images you want, it may just be because your prompts look like this -
Create an image of a black sedan
You can do better than that..
Here’s an example of a slightly longer prompt that gave me almost exactly what I was looking for -
Create an image of a 2000 Toyota Camry sedan. The car should be black with tinted windows. The image should be photorealistic.
Simulate using a Sony A7Rv camera body with a Sony FE 135mm f/1.8 GM lens.
The background should be a mountain with a setting sun.
Longer prompt, yes, but look at this image -
I said that it was almost exactly what I wanted. Ideally, I wanted the background to be blurry. The 135 lens is buttery smooth and it creates a really nice bokeh (blurs the out of focus parts of the image).
I didn’t really ask it to do that though, did I? Let’s fix that
Generate a similar image but this time keep the car in focus and add a strong bokeh behind it
Brilliant!
Yeah, the 135 would produce a better image. This is still pretty damn good.
Back to 1 - if you aren’t getting exactly the images that you want, you aren’t being specific enough with your prompts (seriously, when is the last time you specified a camera and lens?)
Don’t take my word for it, try it out…