I’ll try to explain here, I used GPT via web to refine a prompt for image generation, I reached the perfect end result, in which GPT generated the image I want to generate in my app, then I integrated dalle-3 into my app, using the same prompt the generated image is completely different, what am I doing wrong?
GPT result:
https://imgur.com/5SxoREh
Dall E 3 result:
https://imgur.com/wedhf2x
Isn’t this simply an example of variation ? Eg. the same prompt will almost never give the exact reply.
The GPT image gen and dalle-3 image gen are different tools, with GPT’s being the evolved version of the dalle-3.
When you get an image that you want to remain static, you upload that to where you need it and use that…
If you need something that is repeated in various poses or settings or scenes, you include the image you like as a reference image and let the same model you built it with generate more based on that reference image.
The idea that different Ai models will produce the same image exactly from the same prompt is asking too much for many technical reasons my friend.
I used GPT itself to refine the prompt and make it more suitable for Dalle. It didn’t work, so I gave up and completely removed Dalle from my project. Thanks for the responses.
there’s another valid solution taken from me.
you were given the answers you needed to complete the project you’re working on bro.
“Bro”, I removed dalle from the project, it was released without an image generator.
Dalle couldn’t meet my project’s demands. Even after testing with several prompts, some very simple, it created a completely unrealistic image.
No need to defend the product; I simply decided not to continue with it.
You’re never going to get the exact same image from the exact same prompt ever.
That was the first part of the technical solution to your issue.
The second part was to use the graphic/image that you liked statically, as an image file in your app and to do so with all other things that need static images.
you asked what you were doing wrong to get the same image to display multiple times across your platform and the answer was given.
now you’re moving into closer into territory that works, because static images don’t require an image generator being a part of your project, which was also the solution that I directed you towards with using static images.
It’s just frustrating when folks you give accurate information to, act like you’re speaking a foreign language then fumble along and do exactly as you pointed them to do…
While feeling and portraying that they did the opposite.
In general, Ai is used to create new things. It’s not made to create the same image for everyone everytime and there’s no need for it to do so because once we have an image we can just use it over and over already without burning tokens, compute, and costing the company time and money to generate the image over and over and over and over.
I wish you the best on your project, I hope that the next time you take the time to ask for help, you also take the time to consider the help provided.
You would think there is lots of diversity to be had behind a gpt-image-1 prompt such as:
"Generate photo: a cute kitten"
Enjoy 100 of them - 145MB of almost the same cat.
https://imgur.com/a/100-gpt-image-1-cute-kitten-9BzNaee
Is it with just the simple stuff?
I’m actually amazed because anything ive run twice has quite a bit more variance.