Flux Prompt Styles

This is a collection of image style prompt snippets suitable for use with wildcards in your Flux images. You can use these with any wildcard insertion helper, but it will work best if your wildcard utility understands __insert_file_here__ syntax. I personally use ComfyUI to generate Flux images and with the comfyui-dynamicprompts custom node.

I found myself repeating the same phrases over and over when generating images of different types and some of those styled prompts worked better than others. This is a collection of prompts I have tested with a variety of images and they work well with Flux's prompting style. They are loosely adapted from some common SDXL styles, but altered to fit the full sentence descriptive format that Flux prefers.

There are several styles that I tried to include but failed to get working well enough for this package. In particular, Flux does not respond well to painting styles and barely listens to illustration directions (comic and manga ok, charcoal and pencil sketches not so much). For those styles that didn't work well in prompts, you'll be better off with a LoRA instead. In some cases, it won't hurt to add a LoRA on top of a prompt style to reinforce the visual theme.

I've rendered and labeled a ton of sample images for your reference. So many that I got tired of it and didn't make a full set of styles for every image. In some cases, the style doesn't work well with the image anyway. For example, a long_exposure photographic style is great for action photos of fast cars, but not so useful for portraits of people. The full list of included styles is at the bottom of this description.

When generating the sample images, I started with a base image with no style and kept the same prompt, steps, and CFG setting when adding each style to the end of the prompt as a suffix. If you want to use my workflow that makes this kind of thing easy to do, you can download it here.

If you don't want to use the dynamic prompting tool, you can just download the styles pack and manually paste the text for the style you want directly into your prompt. Then you can have fun experimenting with your own styles and tweaking mine to fit your own tastes. Go crazy. Maybe you can even make "styles/watercolor" work for you because I sure can't.

The base images along with their prompts and settings are below. All of the style samples are posted to a collection that you can find here. Sorry I don't have the time or patience to separately link each style to the associated samples.

• Prompt: A small house stands by a country road. A cat sits on a step by the front door. Rolling hills and forest make up the landscape in the background.

• Size: 1024x1024

• Steps: 25

• Flux Guidance: 2.5

• Seed: 5678

• Prompt: A racecar is driving very fast on a Grand Prix race track in Tokyo.

• Size: 1024x1024

• Steps: 25

• Flux Guidance: 2.5

• Seed: 3456

• Prompt: A woman is sitting on a stool at the counter of a late night diner.

• Size: 1024x1024

• Steps: 25

• Flux Guidance: 2.5

• Seed: 2345

• Prompt: A pretty blonde woman wearing a dress poses standing in her living room.

• Size: 1024x1024

• Steps: 25

• Flux Guidance: 2.5

• Seed: 1234

The full list of styles looks like this (copied from the "random" wildcard file)...

__styles/3d_render__
__styles/anime__
__styles/antique_film_photo__
__styles/architectural__
__styles/biomechanical_cyberpunk__
__styles/biomechanical__
__styles/black_and_white_film__
__styles/clay__
__styles/comic__
__styles/cybernetic__
__styles/cyberpunk__
__styles/digital_art__
__styles/disco__
__styles/fairy_tale__
__styles/fantasy_art__
__styles/futuristic__
__styles/glamour__
__styles/gothic__
__styles/grunge__
__styles/horror__
__styles/line_art__
__styles/long_exposure__
__styles/manga__
__styles/minimalist__
__styles/realistic_film_photo__
__styles/silhouette__
__styles/steampunk__
__styles/tilt_shift__

There is one other style named styles/random which includes all of the above styles. Using the random style will cause the dynamic prompting tool to randomly pick a style for each image.