Dear Moderator

This model does not depicts any actual person:

This model was not trained on real imagery of a living, or deceased, person, or depicts a character portrayed by a real-life actor or actress.

• The face is merged based on multiple faces generated by serval SDXL models.

• The training set are all generated by Flux and SDXL, then use face swap to replace all the faces identically.

• The aim is to train a model that can have consistency over face generation, while all blend in a certain photo style with it.

• Please check dataset if you still have doubt, the face is pretty obviously unrealistic, sometimes even horrible. You can also notice blurry pixels and twisted edges (e.g. walls) in the backgrounds of the dataset images.

Please unlock if you see the message.

Description

Hey there, fellow creative souls! 🌟✨

Introducing Fictional Characters - Crystal Zhao, a tantalizing LoRA that promises to bring your favorite fictional characters to life in a way that's both thrilling and... well, let's just say, a bit unorthodox. 😏✨

Training Data

This LoRA is designed to generate images of a particular fictional character, but there's a twist! It's not your typical character dataset that got formed; it's a fusion of two models working in tandem:

1. Pose Model - This model focuses on creating dynamic and sometimes odd poses, giving your characters a unique stance that you might not expect.

2. Face Model - Here's where the magic happens with SDXL (Stable Diffusion XL). This model crafts the faces, aiming for a blend of realism and the uncanny valley.

Potential Bias

Here are some biases to keep in mind:

- Non-Live Appearance - Due to the dual-model approach, characters might appear less lively or animated than intended, giving them a static, doll-like quality.

- Blurriness and Detail Loss - The merging process can sometimes result in loss of detail or a general blurriness, affecting the overall realism.

- Weird Character Behavior - Since the models generate poses and faces independently, the resulting images might show characters acting in ways that are not typical or expected for them.

- Static Eye Contact - Characters might have a fixed gaze that doesn't track or engage with the viewer, giving them an eerie, non-live appearance.

Remember, this LoRA is still in development, so if you decide to give it a go, brace yourself for some unexpected results! And if you do, I'd love to hear about your experiences and see what you come up with. Let's make some magic together! ✨💖

Usage

Use the following trigger word:

The image is an iPhone raw photo of Chinese influencer Zhao WanTing.