Your AI-generated Art

So I have a ChatGPT "ímage describer" where I feed it a image and it brings back a JSON prompt of said image
@OnlyOneKenobi feed this into the machine and show us what comes out ?
1767281457771.png
{
"prompt": "A solitary hooded figure in a long, dark textured robe stands with head bowed and hands clasped in front, viewed from behind in a vast, barren desert landscape under a moody, overcast sky. Towering, gigantic mushroom-like structures rise around them, with thick, gnarled stems supporting wide, umbrella-shaped caps featuring intricate radial gills underneath, textured like ancient, weathered rock formations in shades of earthy brown and beige. Scattered rocks and debris litter the sandy ground, with a hazy fog diffusing the warm golden light breaking through clouds on the horizon, casting long shadows and rim highlights for a dramatic, cinematic atmosphere.",
"negative_prompt": "blurry, low quality, deformed, ugly, extra limbs, mutated hands, poorly drawn face, bad anatomy, watermark, text, signature, overexposed, underexposed, cartoon, anime, illustration",
"steps": 50,
"sampler_name": "DPM++ 2M Karras",
"width": 1024,
"height": 768,
"cfg_scale": 7,
"seed": -1,
"style": "cinematic, photorealistic, surreal",
"lighting": "volumetric backlighting, golden hour haze, soft shadows",
"quality": "8K, ultra-detailed, sharp focus"
}
Click to expand...
 
So, various image generators have their own style of prompting. That JSON format might work well for Nano Banana or Chat GPT, but I generally use Flux or Z-Image lately which uses more natural language. I took your input image, ran it through my Flux image describer and got the following prompt :

An enigmatic figure in a dark, weathered cloak stands solemnly in a vast alien desert, head bowed beneath the towering presence of colossal mushroom-shaped stone formations. These surreal megastructures stretch into the stormy sky, their wide, gilled caps catching rays of diffused sunlight piercing through the clouds. The landscape is dry, windswept, and ancient, with drifting sand creating a hazy, cinematic atmosphere. Cracked rocks and remnants of eroded stone structures scatter the ground. The figure’s cloak billows subtly in the dusty wind, emphasizing solitude and mystery. Lighting is dramatic with warm golden highlights and deep shadows cast by the towering fungi. The entire scene evokes the feel of a distant, forgotten world — alien yet strangely sacred. Shot in ultra-realistic cinematic detail, using a wide-angle lens, shallow depth of field, and moody volumetric lighting.

Which gives me this image.
FluxExample.jpg


If I just use your JSON prompt, I get something a little less satisfactory (image below), which makes sense considering it's not quite the right syntax for Flux.

JSON.jpg
 
Anthro said:
@OnlyOneKenobi are you running it on your own PC or on a platform (paid for) ?
Click to expand...
Images in this thread, 99% local including the last one. I did use Nano Banana to generate the images to create the training data to make the Lora models for Flux and Z-Image, which I run locally / offline in ComfyUI.
 
Always had a pain getting ComfyUI to run, used InvokeAI mostly - maybe time to get back to understanding how that installs
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X