i tried to creat a book cover for a book i am writeing. but instead of just creating the images to later use for the creating of the cover, it created an image of a book.
the coverst themself created are AMAZING but not to use as a template for the final cover.
and even i wrote in the prompt not to insert the Title or any other Text, there is some Text on the covers, and mostly not really the correct words and only some glyphes that are simillar.
the prompt i used first was:
create a book cover on the topic “Inner Circle Meets Institutions: Master Strategies in Trading”. It should be divided in half into a front and back cover. There should be no title or other text on the cover.
also a modified prompt did not help and dalle just created more amazing but usless images:
create a book cover on the topic “Inner Circle Meets Institutions: Master Strategies in Trading”. It should be divided in half into a front and back cover. It should serve as a template for the book printing. There should be no title or other text on the cover.
the upper two images were made with the first prompt, the lower two images with the second prompt. all would look amazing for the bookcover if they would be in the correct format and the title would be written correct even the font style would be perfect.
DALL-E is notoriously bad at producing text in images from what I’ve seen; that is going to have to be produced in post-processing via Photoshop or your editing tool of choice.
The prompt itself “book cover” is going to push DALL-E to generate images as if they were on a book. My suggestion would simply be to describe the image you’d like to see. You can also specify resolution and dimensions now, so if you know what you’re working with size-wise, you can specify that, dump the image in photoshop, add your text, and bam you’re golden.
As a writer, think of it like describing the scene. It will take your descriptions very literally, unless you use abstract language, like emotive language. Modifying the image to be placed as a book cover will have to be done by you, but you can describe what you want visualized so it can produce that base image for you to modify.
It’s a handy step to add into a larger workflow process that will speed up time, but it’s not a catch-all right now. DALL-E still has a lot of limitations.
I’ve had some success. I had it generate a tall image then added text on my own in Photoshop…
This is a prompt that worked…
Epic and dramatic photo illustration of Leo for a kids chapter book. Leo is a young boy with wide eyes and jet black curls. He wears adventure-ready clothing and is accompanied by Flutter, a vibrant butterfly with a unique wing design. Leo is standing near a frozen lake, with the atmosphere capturing a sense of grandeur and intensity.
You can also play with terms like “Rule of Thirds” to make sure you get a decent “cover” image…
I’ve also found that saying “there is no X” or “without Y” can sometimes inadvertently lead the model to actually adding X or Y to an image, so you may have luck phrasing without those.
