>>43948331>feigned helplessnessSo try to learn
>masterpiece, highest quality, (ceres fauna looking like a goddess:1.7), (walking in a garden:1.4)<lora:allVtubersLora_hll31:1.0>You are using the allVtubersLora_hll31 LoRA at weight 1, with a model that already has hll1 inside it
HoloCreamsicle. I linked you to two models. Download them and put them in your models/Stable-diffusion folder. You can delete "HoloCreamsicle" since it's twice as big - these two together are the same size combined. I only thought you wanted to use both HLL's because of the LoRA, so you can skip "One Last Creamsicle" and just download ReverseCreamsicle
plays better with other LoRA because of the way it was merged, confirmed and vindicated>512x760lol, x768, but that's fine
>masterpieceDon't use it. We're not on NovelAI anymore.
>(ceres fauna looking like a goddess:1.7), (walking in a garden:1.4)Sir, too much. Also, we're not using SD1.5 itself, these text encoders are different.
>Steps: 40, Sampler: Euler a, CFG scale: 10.5, Seed: 2309677027, Size: 512x760, Model hash: acffdca090, Model: HoloCreamsicle.pruned, Clip skip: 2, ENSD: 31337You can get away with using a DPM++ sampler with 20 steps instead of Euler a at 40 steps.
>Negative prompt: (worst quality, low quality:1.4), (depth of field, blurry:1.5), (greyscale, monochrome:1.2), (bad_prompt:0.8), (censored, censorship:1.3), (text, speech bubble), error, bad anatomy, (bad hands), sketch, multiple viewsMatter of preference really but they're probably fine. Unless you don't actually have the bad prompt embedding.
>>43948594>The numbers are something that got stuck onto the filename by default when I pruned it. I have no idea what they mean but felt no need to delete them because they might mean somethingThe model toolkit adds them, yeah. I only keep them at the ends of mine if I've deliberately done something like swap out the text encoder, which is the last of three numbers. VAE-v1-SD, CLIP-v1-SD, UNET-v1-SD.
