AI ART: news & 🐙🐝🍆 spaceship fluff (Patreon)

HASSANBLEND1.5 RELEASED

A newer version of the awesome HassanBlend AI model for StableDiffusion has been released yesterday! You can download it from here:
https://huggingface.co/hassanblend/HassanBlend1.5/tree/main . You only need the .ckpt file and place it in models/stable-diffusion.

You can find more info about HassanBlend, tools to make better images with it and if you want to show support to the dude that trained the model here: https://www.patreon.com/sd_hassan

I haven't tested it a lot yet but after a few quick tests it looks awesome. It's much more body oriented and the AI locks in more into the results rather than going random with it.

v1.4 example 

v1.5 example for the same prompt

prompt was "photo of sexy blonde woman with perfect body in scifi suit, blue eyes, gorgeous, pouty lips, makeup, stars in the background"

SPACESHIPS FLUFF & TRICK

I was testing how good AI generated images can get for scifi/alien images. The video above is with some of my earlier tests.

The prompt was something like "photo of (alien spaceship) floating in space, (scifi ship), stars, planet in the background, mechanical, futuristic, robotic, octopus, underwater, highly detailed, photorealistic" and I used HassanBlend1.4 as the model. First 10 images were with "octopus" thrown in there to make the ships go more towards that direction/shape. Then I replaced that with "bee" and finally with "eggplant" because why not.

I've seen people try to do AI spaceships on reddit but they often end up very boring and human-like, for example:

The trick to get more interesting sci-fi images I think is to throw some random words in there and then try to modify the prompt so that the AI doesn't go too crazy with it. If you add "spaceship,octopus" the AI will do actual octopuses in space. By emphasizing "robotic, mechanical, ship", etc. like I did, the AI will gravitate more towards that side of things, while the random element will get incorporated more subtly: