I recently had time to test some GPTs in the store, choosing the dalle group because it wasn’t very complicated to test. Selected from high, middle, and low ranks. of another acquaintance
I closed the name GPT In order to avoid damage to the model used for testing.
The first time, I threw in a single prompt for all of them, and the result was that all of the images had similar properties. It is considered to be lacking in the basics of creating a logo and for creating a tattoo style. I know more about the feeling than I have yet to pass in terms of a tatto.
The prompt created is a background prompt used to send to DaLLE. Even though it’s different in GPT’s settings, the one created is a broad functionality that relates to documents (trand 2024, not related to tattoos or creation methods). specific instead shows images that are more tattoo-like. There is also a role-based test of GPT that concludes from their behavior that if it were me, I wouldn’t think of doing it. Because it is a waste of user messages. Contrast that with me creating a GPT that can generate the first image in 2 messages and using few tokens. Perhaps gaining popularity takes more than just effective work