How to generate full body images?

I have a very detailed description of a person for generating photo’s. Description is about the face and full body. Each time I use the prompt Dalle generates ‘more or less’ the same person. But it is really difficult to generate full body photo’s of this person. Every time the person is generated close up. I tried a lot of things like "use wide angle’, ‘show person in background’ or ‘make sure the full body of the person is visible’. But is seems really difficult to let Dalle understand this part of the prompt. Someone any idea’s?

It’s about the prompt. I’ll share mine with you and upload a photo so you can see what I mean.

In an inimitable fusion of art that capture the Platonic/Socratic Trinity. Use a heavily textured style, use impasto and chiaroscuro, with brushstokes of thickly laid-on paint to recreate a Post-expressionist painting that blends Van Gogh and the Mosaics from the Hagia Sophia: capture three feminine figures in their elegance: powerful, dramatic, lyrically rhythmic, imaginative, and emotional.

Begin with a base of textured impasto, layering thick, vibrant strokes of oil paint to create a dynamic, pulsating background that evokes the emotional intensity and rugged physicality of Van Gogh’s Post-expressionist works. Within this tempestuous sea of color and form, employ chiaroscuro to sculpt depth and volume, casting shadows and light to carve out the painting.

Let the abstract interplay of light and shadow, form and void, represent the three female figures. Imagine them as a radiant from which forms emanate and to which they return.

Incorporate elements reminiscent of the Mosaic of the Transfiguration, using tesserae-like strokes to imbue the piece with a sacred, timeless quality, as if each brushstroke is a fragment of a greater, divine mosaic. These fragments should not form coherent figures but should hint at the underlying unity…

As you paint, let your brushstrokes dance between the realms of the material and the metaphysical, creating a composition that is both powerful and dramatic, yet suffused with a lyrical rhythm that speaks to the deepest intuitions of the soul. The final artwork should be a testament to the imaginative and emotional depths of human experience, transcending the limitations of conventional representation to touch upon the ineffable mystery of existence itself.

1 Like

Image generation is by far the least developed of the GPT’s current functions, in my opinion. I have been able to generate more-or-less full-body images…usually just by being direct in conversation with the GPT as if I were talking to an actual graphics artist. But there’s still much left to be desired. I once asked it to generate a girl wearing flip flops…and so she was…but her feet (in sandals) were right where her hands should be…and she had seven toes or something too. Slightly creepy, even with a nice manicure…I mean pedicure. Another time I was trying to get it to make a generated custom GPT image (simply the face) stay the same and look a little older, and it came up with three faces side-by-side instead of one.

But if all you’re looking for is a full body shot, I’m pretty sure I could sweet-talk it into that if you gave me the details.

Hi, I’ve been trying out DALL-E 3 for about 2 months now (on Windows 10 laptop computer using the Image Creator from Microsoft Designer on Chrome), and I have experienced the same trouble of getting a full body image on-demand.

However, I have found that if I specify certain poses or activities that require the whole body to be shown in order to actually portray it, then I have more success in generating full body images.

I have some examples, though not perfect images, that do show the whole body I’ve tested just today. I’m a new user on this forum, so I can only post one media picture at a time in my post.

woman sitting cross-legged on the grass in a woody park



Other prompts I did to get the whole body were:
woman doing push-ups in a gym

woman ballet dancer on her tip toes elegantly posed

woman in street clothes sitting on the hood of a car in a new york city neighborhood side street

I’m still personally trying to figure out how to more consistently get full body images of characters/people, but for now, this is my workaround. I hope this helps!

1 Like

The challenge is more with the tall aspect ratio size. Even including it in the prompt. You can see the AI composed the starting square so it could be expanded to be tall, but the drawing rotation was wrong.

Compensating for the problem by informing DALL-E of the rotation needed:

“The tall aspect ratio photo is rotated 90 degrees to show full body length, showing the person head to toe.”

(The exact same prompt with simply “female” instead of “male” is refused over and over, not by content policy, but perhaps based on vision)