DALL·E 3 Image prompts

Sioux Brave
Sioux Brave
sioux brave. Special Effects: Fractalius
Generated by AI Power using DALL-E 3 Revised prompt: A Sioux warrior standing tall in traditional attire. His features are being distorted and made surreal by a Fractalius-like special effect. This special effect visually resembles the intricate geometry of fractals, highlighting and amplifying patterns and colors across the image for a unique aesthetic. The warrior’s attire, perhaps a headdress, war paint and beaded clothing, are all morphed by this effect into an explosion of colors and shapes that evoke a feeling of otherworldliness.
1024 x 1024

Do you want to know how to write good image prompts for DALL·E 3?

Yes, so do I!

This isn’t exactly a guide or a cheat sheet but it does record some observations.

The cheat sheet is here: DALL·E 3 image generator prompts crib sheet

AI Power > Image Generator

Using AI Power‘s Image Generator, I traced its request to OpenAI to produce a DALL-E 3 image, Type: Natural. I wanted to see how it applied different Settings to the Prompt that I’d written.

sioux brave

I chose Special effects: Fractalius

The prompt sent to AI was

sioux brave. Special Effects: Fractalius

The revised_prompt, returned with the generated image, was

A Sioux warrior standing tall in traditional attire. His features are being distorted and made surreal by a Fractalius-like special effect. This special effect visually resembles the intricate geometry of fractals, highlighting and amplifying patterns and colors across the image for a unique aesthetic. The warrior’s attire, perhaps a headdress, war paint and beaded clothing, are all morphed by this effect into an explosion of colors and shapes that evoke a feeling of otherworldliness.

It’s this revised prompt that’s passed to the image generating engine. The result is the featured image.

oik-ai: my AI tool

I repeated the experiment with my AI tool, which uses Type: Vivid. This time the revised prompt was

A portrait of a brave indigenous person from the North American Sioux tribe. The individual carries a dignified and stoic look, displaying the features commonly found within their community. The image is transformed with fractalius effects, diffusing the edges into an intricate pattern of vein-like structures, adding an otherworldly aura that makes the image look like a living piece of stained glass art.

Then I tried entering the revised prompt that I’d got using AI Power. The new revised prompt was

Describe an image featuring a Sioux warrior standing tall, adorned in traditional attire. He appears amidst an intricate geometry of unusual aesthetics, resembling fractals. Surrounded by an explosion of colors and shapes, his attire – including headdress, war paint, and beaded clothing – is enhanced and morphed by this surreal distortion. The overall effect amplifies all elements in the scene to evoke a feeling of otherworldliness, transforming the warrior’s features and attire into a mesmerizing visual journey.

And because it’s AI, working on a different prompt, I got a completely different image.

Methinks it’s time for a rephrase of a well known saying

Insanity is doing the same thing over and over again and expecting different results. 

usually attributed to Albert Einstein

Is every image unique?

As you can see, every image is different. But what if you wanted to create a series of AI generated images with the same theme, eg for a children’s book?

Well, on the OpenAI forum, there are some suggestions of how to do it using DALLE-3 in ChatGPT.

There are a couple of tricks.

  1. Tell the AI tool not to fiddle with the prompt.
  2. Ask for a grid of images, and then use an upscale tool.
  3. Use DALL-E in ChatGPT 4, with Custom Instructions.

Trick 1: Don’t fiddle with the prompt

The documentation for DALLE-3 says use this prompt

I NEED to test how the tool works with extremely simple prompts. DO NOT add any detail, just use it AS-IS:

The revised prompt for original message ( Sioux Brave. Special effects: Fractalius ) was

A Sioux brave with Fractalius special effects

In the OpenAI forum there are suggestions that you can use different versions of this instruction.

My prompt has full detail so no need to add more: … your prompt here …

When I used this trick the revised prompt was different from the original.

A Sioux warrior standing tall in traditional attire. His prominent features are being distorted and molded into a surreal form through a special effect that shares similarity with the complex geometry of fractals. Such transformation of the image outlines and amplifies various patterns and hues scattered across for a distinctive aesthetic presentation. The traditional wear of the warrior, possibly a headdress, war paint and beaded clothing, are all reshaped by this effect into a vibrant burst of colors and patterns triggering a sense of everything being from another world.

Another method ( see the references ) suggested changing the last sentence in the prompt. It also added. n=1. Do not alter the prompt.

“Sammy is a playful bear with wide eyes and jet black curls. Buster, a distinct robot, accompanies him. This is a landscape-oriented single photo illustration in b&w for a kids chapter book interio . Sammy and Buster are in a swamp.” n=1 Do not alter the prompt.

Well I tried that prompt AS-IS, with the typo for interior, and got something pretty similar. Then I tried my own version

“Sammy is a playful bear with wide eyes and jet black curls. Buster, a distinct robot, accompanies him. This is a landscape-oriented single photo illustration in b&w for a kids chapter book interio . Sammy and Buster are with a Sioux Brave.” n=1 Do not alter the prompt.

Trick 2: Ask for a grid of images

sioux brave. Create a grid of 4 images of the same sioux brave, riding a horse, making smoke signals, hunting buffalo and standing by his teepee. Special effects: Fractalius

The revised prompt was

A quartet of images showcasing a Sioux male warrior engaged in different activities. In the first square, we see him riding a horse across a vast plain, the wind sweeping through his hair. The second square captures him atop a high ridge, conjuring smoke signals into the sky, the smoke spiralling and twirling in the air. The third demonstrates him in hunting attire, pursuing a buffalo with a bow and arrow, poised and strategic. The final image presents him standing stoically next to his teepee, a regal and serene stature. Each image shall have a unique fractal-inspired effect, giving dimensions of abstraction and surrealism to each scenario.

Trick 3: Custom instructions

To be completed! I haven’t used ChatGPT 4 yet.

References