>>18528130Hi other british poster!!! The best site in the OP is nemusona, you should use that. Or the google colab, but the guide on it in the OP is currently broken. Anyway something to note is that the negative prompt is just as important as the positive prompt, you should just use a catch-all like this in the negative prompt field:
>(worst quality, low quality:1.3), extra fingers, fewer fingers, lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist nameFor a boilerplate positive prompt you should put smt like this at the start no matter what the rest of the prompt is
>((8k, masterpiece, best quality)), ultra high res, absurdes, BREAK, ...Then I would recommend going to danbooru and copying the tags of an image you like. Nemusona supports this through the Import Tags button, but I find the copy all danbooru tags to clipboard to be easier. Keep in mind, stable-diffusion is not good at generating more than one different character, it will blend the tags together if you describe 2girls or even just something like a red triangle on top of a blue sqaure. And also, copying tags alone will usually not generate specific girls properly unless they are uber popular like e.g. Remilia Scarlet who has 48k images on danbooru, in that case just typing (((remilia scarlet))) will generate her--the parentheses indicate strength. Otherwise, if you have a character in mind you should scour danbooru for tags that she is usually described with. In my case, these are the tags I use for Nishizumi Miho from Girls und Panzer
>1girl, solo, (((nishizumi miho))), ((girls und panzer)), short hair, brown hair, brown eyes, medium breasts, skinny, bangs, hair down, straight hair, (tareme), ......and then I add the rest of the prompt which you can just copy from a danbooru img you like!
