Creations by GravidaSomnia

()

()

Someone asked about my avatar, so I played around with the concept some more using more recent AI model blends. I can't re-create the original but the new ones aren't bad either.

()

preggohotties · (4107)

Love the popped out belly button in the pics. Keep it up, love your work.

trmbnlvr · (282)

Wonderful stuff.

()

Thanks!

()

azarafa · (534)

wow, they are so beautiful. especially the redheads.if you could try making a bit more.

()

Maybe later. It's not that hard but the redheads needed a bit of work to not end up with too much weirdness going on with their bellies; I tend to just keep moving on to new things to test out.

It's worth having a try at making them yourself. Especially with the newer AI models that have been released, it's easier than ever to get decent pics.

Here's the settings I used for txt2img, everything except the random seed and AI model:

Quote:Prompt: 400 ISO, f/4 aperture, 8k HDR, sharp focus, award winning RAW color photo of a young pregify woman, italian irish redhead, nude in the parking lot, ((opening a long thin jacket)), 1992, pregnancy, ((pale)) freckled body, curvy and thick

Negative prompt: muscular, strong, skinny, thin, lowres, poor anatomy, text, error, glitch, cropped, worst quality, low quality, normal quality, jpeg, signature, watermark, username, blurry, out of frame, bad art, deformed, disproportionate, blurred, blur, mutated, pixel, monochrome, b&w, 3d render, long neck, morbid, photoshop, video game, 3d render, oversaturated, anime, cartoon, surreal, grain, decayed, greyscale, amputated, amputee, same, tanned

Steps: 24, Sampler: DPM++ 2S a, CFG scale: 6.5, Face restoration: GFPGAN, Size: 640x720

I'm using a blend of photorealistic models from CivitAI plus LadyBabyloin's fine tuned model available in their thread.

I mostly use the Automatic1111 WebUI implementation of Stable Diffusion but there's other options.

To upscale I use img2img and up the resolution (same prompt usually works but sometimes it needs editing):

Quote:Steps: 16, Sampler: DPM++ SDE, CFG scale: 7, Seed: 2621205191, Face restoration: GFPGAN, Size: 1024x1280, Denoising strength: 0.28, Mask blur: 4

I tend to work on one pic at a time, other people seem to prefer running batches and picking out their favourites. Some fine tuning of the settings and editing is usually required when aiming for photorealism, both before and after upscaling. But the results are worth it imo.

Artistic styles are far more forgiving and produce good quality pics more consistently.

Anyway, here's a couple of pics for today.

()

last posts for a while.