I run into this issue with various models. It helps to put 'frames, framed, border, cropped, 2x2, 4x4, 6x6' in your negative prompts

that's likely because you generate square images. Try generating in 512x768

Try a different seed :D

@Cyberdelia looks pretty cool... I'll be trying some of the Bionic Studios's TI with this model (:

@Cyberdelia (Sorry for the late rep,been really busy IRL.)

First of all, oh no, thank YOU, good sir/miss, for your service to the Community. CyberRealistic is my reigning Photorealism Model. It used to be RV--and it's still good when I'm just RNG-ing it (ala MidJourney), but as I mentioned on another post, I just got tired of its stubbornness. So it got toppled with URPM, which is VERY obedient, AND it's awesome even for SFW stuff, but a bit lacking in the fantasy & sci-fi departments (& I don't blame it,it's just not its specialty;same w/ Chillout). Then you rolled in with CR--or at least a juiced-up version of it (I'm very new,just weeks-old,so I wasn't there for your OG editions), and yup, it did it. There are other awesome PR models out there, but they're my secondaries, and this is my primary/mainstay for these kinds of projects.

As for the image, yes! I just saw it on the front page. Thank you for bringing it back, author-san. As I've said, you did NOTHING wrong. But if Civitai never gave you a reason/strike to begin with, it just... "disappeared" it, then wouldn't it be just a matter of time before it hap-- static,then silence...

I've been using this model since 1.3 and it's my favorite model. There are some other good ones for different styles but I keep finding my way back to this one.

for the hands i negative with: .. ,mutated hands and fingers, disconnected hands, .. nothing more. No TI or other things. Look to keep under the 75 token barrier, it helps to respect all "whishes" :-) .. 3.0 is a little bit sensitive on denoiser and cfg-s over 9.0 ..

@joge25 ok thanks i will try, cause i really like some colors of this model.. what you mean with 75 token barrier?

To have a good balanced and respected prompt-set i give each prompt a chance, by limit the amount of prompts and strength. Many prompt-sets later i stopped to use attentions and emphasis, and the most importants are at the start of the prompt-flow. So i have on each run / batch the result which i prompted, and limit the amount of random. If you "break" the 75 Token, then it expands to 150 token, which costs transformer-time and attention. Together with the sample-steps, the cfg and may some Lora or TI attention, it was mostly more "luck" than "plan" on my results.

--

Typing past standard 75 tokens that Stable Diffusion usually accepts increases prompt size limit from 75 to 150. Typing past that increases prompt size further. This is done by breaking the prompt into chunks of 75 tokens, processing each independently using CLIP's Transformers neural network, and then concatenating the result before feeding into the next component of stable diffusion, the Unet.

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#infinite-prompt-length

I've had much better luck with my negative prompt below. "bad-artist" is a textual inversion you'll need to download. It's a bit excessive in length and may not need to be this long, but it's become a stable that I tweak minimally for every image. Also, unless you are using loras or textural inversions in your positive prompt, don't go any lower than 80 steps, set the CFG to 3.5, use a max resolution of 512px in any dimension, and use some sort of "hires.fix" to scale up from there. If you're GPU doesn't have the VRAM for what you want, then, in img2img, use the "Ultimate SD Upscale" extension (Look up a video tutorial on YouTube.) to scale up the image you generated earlier.

Negative Prompt:

(bad-artist:0.7), (worst quality, low quality:1.4), bad anatomy, deformed, (mutated body parts), underboob, (bad quality), seams, text, signature, watermark, distortion, lowres, bad hands, error, missing fingers, extra digit, fewer digits, cropped, jpeg artifacts, username, blurry, mutation, extra limbs, disfigured, out of frame, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, mutated hands, fused fingers, too many fingers, long neck, multiple heads, monochrome, bad feet, ugly, duplicate, morbid, mutilated, unclear eyes, low quality body parts, missing body parts, disproportional body parts, indistinguishable body parts, branched body parts, rheumatism fingers, ugly fingers, melted fingers, less details, censored, (futa:1.1), time-lapse, distorted, unnatural, unrealistic, broken, warped, mutation, ugly, bad hands, contortion,

10000000000% agree it !!!!!

agree!!!

@Cyberdelia there are deficiencies, especially in the vaginal area and sex positions when using Lora