@Background Pony #539D
Thank you, this is an interesting way, but in my opinion openai has crossed all the lines with such a lot of censorship in the model. In the case of generation via bing, there seem to be 2 censoring neural networks: the first checks prompt, the second checks the result. In fact, the censorship tool is still being trained on us. But the model turned out to be too good to abandon it

@Background Pony #539D
I simply specified in the request that it generate a 16:9 image

@Background Pony #539D
That’s a really interesting way to go about it! It definitely worked out well here

@Background Pony #539D
How’d you get it to generate widescreen like that?

I am very glad to know that you also liked it. I asked prompts in DALL-E 3 not directly, but via chat.openai.com, they were pre-processed by GPT4. The main problem when generating via DALL-E 3 is their content policy. Because of this policy, I can’t directly tell him to draw Princess Luna or draw an MLP-style pony. I tried to effectively describe the appearance of Princess Luna and somehow hint to him without a direct text what kind of character I want to generate.

As a result, I got this prompt for GPT4:
“Can you draw a dark blue alicorn? Indicate in the prompt that this is the princess of the night. The alicorn should have a stellar and ephemeral tail and mane, a small black tiara in the mane and a uniform massive black necklace with a painted white crescent on the chest. Her anatomy should be perfect, she should look graceful, the anatomy should be more “cartoon” than like a real horse. On her rump is her “cutie mark” - a dark cloud surrounding a crescent moon. She has big eyes and they look into the camera picture format 16:9. Landscape - a sleeping medieval city, it stands on the pavement”

GPT4 generated this prompt to generate this image in DALL-E 3:
“Illustration of a graceful dark blue alicorn with a starry, ephemeral mane and tail. She wears a black tiara in her mane and a bold black necklace with a white crescent depicted on her chest. Her cartoon-like anatomy distinguishes her from a real horse, and her cutie mark features a dark cloud with a crescent moon. Set against the backdrop of a sleeping medieval city on a cobblestone pavement, she looks intently into the camera.”

I also made small changes through photoshop in the area of the cutie mark and nose.

cool style

Would like to know how you did this. Would think it was legit art if it wasn’t for the bench legs. You did well.

@Background Pony #539D
+1 still trying to understand how..

@Background Pony #539D
Seconded.

This has got to be the best AI generated piece of art I’ve ever seen. Hands down.