Since last time, I've switched over the the stable diffusion webui (Installation instructions for windows: https://stable-diffusion-art.com/install-windows/) My initial testing with the standard SD model generally didn't produce particularly good results. This changed dramatically with the model at https://civitai.com/models/120033/chimeramixl. Just from some very brief testing, it easily produced these images. The chimera model has been trained on a lot of fantasy creatures and knows what a vulpera or a worgen from world of warcraft (for example) looks like. A prompt like "A translucent red balloon shaped like a female vulpera" produces very good results. You have to tinker with the samplers a bit for best results (Euler works really well.)
More Stable Diffusion Testing
Collapse
X
Comment