Why images are too different?

I tried a lot generate an imagem from DALLe3 API near the ChatGPT images, but i never reach it, and it is so frustrating, because our system call it API to generate storybooks.

I use the same ChatGPT prompt and the results are too different/poor:

A cartoon-style image of a historical king inspired by the biblical figure of David, but not depicting any real person. The king has olive skin, dark brown hair, and is wearing a regal, ancient Middle-Eastern outfit typical of the era, complete with a crown and a lyre in his hand, symbolizing his attributes as a musician and a warrior. He stands in a pastoral landscape that could resemble the ancient hills of Judea, with a city in the far background. His expression is thoughtful and noble, as he looks towards the horizon.

And the results:

I put more data here:

How i can get the same/near results from ChatGPT, but by API?


1 Like

The API is currently DALL-E 2. ChatGPT has had DALL-E 3.

When the new model choices are available in API as an API parameter, you’ll have to specify the model you want to use - and thus the price you will pay per image.

Model Resolution Price
DALL·E 3 1024×1024 $0.04 / image
1024×1792, 1792×1024 $0.08 / image
DALL·E 3 HD 1024×1024 $0.08 / image
1024×1792, 1792×1024 $0.12 / image
DALL·E 2 1024×1024 $0.02 / image
512×512 $0.018 / image
256×256 $0.016 / image

“HD” are described as having enhanced detail, likely another round of transformer-based upsampling.

DALLE3 is not available via API yet… about another hour…

New docs here…


Looks like they will be changing prompts…

When you send a generation request to DALL·E 3, we will automatically re-write it for safety reasons, and to add more detail (because more detailed prompts generally result in higher quality images).

Im using python-openai from github and i think that is dall-e 3.

Now, checking their code i think they use azure or other thing.