Introducing Shannon for Z-Image-Turbo and Flux.

Shannon was first released for Wan and now comes to Flux for still image generation. She is a slim, fair-skinned redhead with freckles, green eyes, and full lips.

Important: This model is 100% fictitious and is not based on images of any real person. She originated with images generated from still image checkpoints and has evolved from there using only AI rendering. No images (neither actual nor AI-generated) of any real person was ever involved in the making of this model.

Prompting hints:

In Z-image, no special descriptions are needed. Shannon will almost always come complete with her fiery red hair, green eyes, and full lips. A standard z-image workflow should work well, and feel free to experiment with your favorite sampler/scheduler combo. I rendered all the samples at 20 steps without any z-image LoRA, just the shannon LoRA by itself.

I get my best z-image results after following guidance from this article by Ggrue. I've been using the dpmpp_2s_ancestral scheduler combined with the "FlowMatchEulerDiscreteScheduler" scheduler after installing the custom node via ComfyUI Manager. A search for "EulerDiscreteScheduler" in the Manager installation tool got me the custom node and installation was easy. After the install, the new scheduler shows up in my ksampler scheduler list with all the others, so it's simple to use.

The keyword is "shannon", but it's optional. The model responds well without it. She should generally appear with long, wavy/curly red hair and green eyes, but flux often has its own ideas about hair and eye color so it usually helps to reinforce that in your prompt. A good starting prompt is something like "shannon. a portrait of a woman with copper-red hair and clear green eyes." The sample photos included several hair styles, so you can prompt for things like: "falling in silky waves over her shoulder", "partly covering her face", "wild and messy", "tied in a messy ponytail", "in a long braid", etc. Mix and match as you like. My favorite look is "her long curly hair is a little messy, tied in a loose ponytail".

She looks good with CFG values ranging from 2 all the way to 10, although 10 is kind of overboard and likely to give you a shiny flux face. I think the model looks best around 2.5 or 3.0, but your tastes may differ.

All of my samples were rendered with strength of 1.0 and no other LoRAs in place, using the standard flux1-dev-fp8 checkpoint with dpm++ 2m sampler and beta scheduler.

I hope you enjoy creating pictures with Shannon. If you do, please post them here or tag them with this LoRA when you post so I can see what people are doing with the model. Enjoy!