STOP - Before you complain about how your generations don't look anything like mine, read carefully.

1) We probably have completely different setups. I'm running A1111 WebUI on an RTX 2070 with Pytorch 2.0 in Ubuntu with the following startup parameters -

export COMMANDLINE_ARGS="--opt-sdp-attention --opt-channelslast --disable-nan-check --medvram --lowvram"

export PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512

Also, my config file is probably much different from yours.

2) There's this blue arrow button to automatically organize your prompt. You DO NOT have to manually configure the prompt. Just copy the prompt, paste it into the prompt field, and click the blue arrow that I've outlined in red.

3) If you run on ComfyUI, your generations won't look the same, even with the same seed and proper settings because ComfyUI uses the CPU to generate noise, while A1111 uses the GPU to do so.

Use nsfw in the negative prompt to block nsfw outputs.

I recommend making a style prompt and adding the following information to it, then using it in most cases with this merge.

8k, 4k, (highres:1.1), best quality;
Negative prompt: (worst quality, low quality:1.4), 3d, cartoon, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name, nsfw

Credits:

https://civitai.com/models/3450/moistmix

https://civitai.com/models/4201/realistic-vision-v13

https://civitai.com/models/12124/hans-rudi-giger-style

https://civitai.com/models/5267/monster-diffusion

Interested in joining the community or supporting us financially to bring you more awesome content in the future? Links to Discord and Patreon are in the profile card on the right side of the screen :)