rough cave
has anyone ever come across a concept that dall-e just cannot comprehend? I've been trying to get it to do Minos Prime, and it does everything so well.... until it gets to his crown-head, where it constantly makes it look like a pinecone.
I've tried everything. I've given it several reference images - front, back, side - and it just cannot do it. I've even gone into detail about how its arranged, even in comparison to its previous attempts.
rough cave
I've been trying again and still can't for the life of me get it to generate it right 😭
lyric holly
I've been trying again and still can't for the life of me get it to generate it ...
New image model dropped last night. Perhaps give it another try?
rough cave
New image model dropped last night. Perhaps give it another try?
😄 That was my first thought when I got the ping in #announcements . Has it definitely been rolled out to everyone now? I'm unsure of how to tell what model GPT is using when image generating
lyric holly
Look for pastel color swirls while creating image.
rough cave
Look for pastel color swirls while creating image.
at this point?
Doesn't seem to be any notable pastel swirls here either
the image it produced is beautiful though, and it actually got the head much more accurate
lyric holly
Doesn't seem to be any notable pastel swirls here either
I think you have a the new model. The initial stage is where I see the pastel, but the second stage looks right.
rough cave
Hopefully I do 😄 I'd love for this generations to be this beautiful going forward. Thank you for your help!
lyric holly
It also has the cinematic realism look, which this new model seems to be tuned for.
Have fun!
rough cave
It also has the cinematic realism look, which this new model seems to be tuned f...
Oh - one last question before you go: does GPT have knowledge/context of prior images it generates in a conversation? Or can that only be achieved by reposting its generation as an attachment / by clicking on the image for the unique "edit" UI? I've been wondering if theres any benefit to doing this or whether I should try to minimise my token usage in a conversation.