What is this?

A big set of tested values for different camera and color settings. Intended to give a big variety of results simply by adding a few words at the end of your prompt.

It gives you a huge number of possible image variations on top of your prompt. Works great both on characters and scenery.

How to use it?

You will need the wildcards extension. Extract the zip file to your wildcards folder (Path: stable-diffusion-webui-master\extensions\stable-diffusion-webui-wildcards\wildcards)

Positive prompt (full usage)

After "(your subject), (your style tags)", add:

(__peaksel_camerashotype__, __peaksel_camerashotangle__, __peaksel_camerabrand__, __peaksel_camerafilmsize__, __peaksel_camerafocus__, __peaksel_cameralensfocallength__), (__peaksel_colorscheme__, __peaksel_colortone__)

Random value from each of the lists will produce a result that may be this:

(gregnobodysd15 in a white shirt), (realistic, best quality, photo-realistic, best quality:1.3), (Medium shot, Eye Level Shot, shot by Leica camera, Shot on 8mm film, Split Diopter, 15mm lens), (Blush Pink, Champagne, and Sage Green colors, Soft colors)

Set your batch count to 20 and watch the show.

Bonus: Ultimate negative

Unfortunately, the wildcards extension is not working in the negative prompt, so feel free to copy-paste the content of "peaksel_negative.txt" directly.

Can be used on both characters and scenery, SFW and NSFW. You may need to add model-specific tags that fix the bugs if any, and (NSFW, ????, ?????) if you are generating SFW characters.

This negative is on the mission to become the most complete one, so feel free to suggest important words if you see them missing. Not intended for anime models.

Your Negative prompt structure

(negative1, negative2), (peaksel_negative content), (model-specific tags: if needed), (SFW/NSFW tags: optional)

peaksel_negative.txt:

(out of the frame, out of the image, out of focus, canvas frame, cropped, duplicate), (watermark, signature, artist name, text, calligraphy, logo, censored, label, jpeg artifacts, username, software, UI, user interface, GUI, scribbles, error), (ugly, disgusting, blurry, poorly drawn, bad art, amateur, lowres), (worst quality, low quality, low rated, mediocre, bad composition:1.7), (deformed, distorted, disfigured:1.3), (bad ???????, wrong ???????, extra limb, missing limb, floating limbs), (mutated hands and fingers:1.4), (disconnected limbs, mutation, mutated, amputation:1.4), (conjoined fingers, deformed fingers, ugly eyes, imperfect eyes, skewed eyes, unnatural face, unnatural body), artist atelier

About the Library

All of these values are collected or generated by ChatGPt and then tested multiple times to make sure Stable Diffusion understands them and the results are distinctive enough from others. Boring results have been removed, so the lists are focused on quality rather than quantity.

Can add unexpected results to your mature prompt in text2image when hunting a good composition OR can be used for quick subject testing (like character embedding for example).

Have fun using it.