Image Composition wildcards & templates版本v1.0 (ID: 159208)

This is intended as an all-around Wildcard list for Dynamic Prompt, that can be used as a template to engineer the prompt for composition's sake, or generate random images.

The wildcards are seperated as follow :

Character

• Description : This include basic description (Age, shape, quantity, hair, eyes, skin, expression)

• Outfit : Clothes and accessories (type, color, noun)

• Pose : Stance /action of the character (pretty basic, without bg impact for now)

• Type : Related to the specy, specific Lora and modifiers

Background

• Sky

• Weather

• Time of the day

• Biome /landscape

• Structures

• Architecture type

Scene

• Theme

• Style

• Colors

• Camera angle

• Artist

All of this can be used to alter the composition of the generated image, and was intended to facilitate random wallpaper generation.

I have regrouped those into 3 differents styles template, which contain all of those wildcards. Note that some wildcards counteract eachothers, so try out what interest you if you have a specific goal in mind. The styles are as follow :

Character Template :

__Composition_My/Character/Descriptors/Quantity__, __Composition_My/Character/Descriptors/Age__ __Composition_My/Character/Descriptors/Bodyshape__ {__Composition_My/Character/Descriptors/Skin_Color_Human__ | __Composition_My/Character/Descriptors/Skin_Color_Other__} colored __Composition_My/Character/Type_Random__,

__Composition_My/Character/Descriptors/Hair_Lenght__ __Composition_My/Character/Descriptors/Hair_Color__ __Composition_My/Character/Descriptors/Hair_Cut__,

__Composition_My/Character/Descriptors/Eye_Color__ eyes, __Composition_My/Character/Descriptors/Expression__ expression,

wearing __Composition_My/Character/Outfit_Random__, __Composition_My/Character/Pose_Random__

Background Template :

{inside|outside} __Composition_My/Background/Architecture_Type__ __Composition_My/Background/Structure__ in __Composition_My/Background/Biome__, __Composition_My/Background/Weather__ __Composition_My/Background/Sky__ __Composition_My/Background/Time__

Scene Template :

__Composition_My/Scene_Modifiers/Theme__, __Composition_My/Scene_Modifiers/Style__, __Composition_My/Scene_Modifiers/Color_Palette__, __Composition_My/Scene_Modifiers/Camera_Angle__, __Composition_My/Scene_Modifiers/Artist__

I used those alongside with the following prompts :

Positive quality :

(best quality, highres, 4k, 8k, uhd, highly detailed, sharp, masterpiece, <lora:Add_Detail:0.7>)

Negative quality :

( FastNegative_v2, BadDream, Bad_Hands_v5), (lowres, low quality, out of frame, cropped, blurry, artifacts, text, watermark, logo, signature, username, deformed, bad anatomy, disfigured, poorly drawn face, bad face, mutation, extra_limb, ugly, poorly drawn hands, bad hand, fused fingers, long neck, bad proportions, extra digit, fewer digits, bad eyes, poorly drawn eyes, copyright, compression, artifacts)

The images in this post are all generated with the same prompt, regrouping all the Styles up here, with mostly those parameters :

• Sampling method : DPM++ 2M Karras

• Sampling steps : 35

• Hires fix : 4x UltraSharp

• Hires steps : 15

• Denoising strentgh : 0.4

• CFG Scale : 7

For the models, I mostly use Deliberate, Ghostmix and Dreamshaper, with klF8Anime2 VAE.

I also used negative embedding such as FastNegative, BadDream and Bad_Hands, and the Lora Detail Tweaker.

Finally, some wildcards (such as Artists) are based on already existing waldcards that I think were included in the Dynamic Prompt extension.

Concerning the characters, you will find a "Lora" file that you can modify for your needs.

This work can very probably be expanded (For example : by adding a "Vegetation" wildcard for the Background), feel free to post your recommandations !

描述:

训练词语: