Character List + AI Models

Preamble

I am planning to train image and video LoRAs (Low-Rank Adaptations) for each of the female characters in my story thread. For now, the image LoRAs will be trained using Flux Dev, while the video LoRAs will be trained using Wan 2.1.

This approach offers significant benefits for my storytelling. For instance, if I want to include a scene where Rias is about to leave the Student Council Room, but pauses before she does, looking over her shoulder to give Sona a soft smile—I can generate that exact image using Flux and seamlessly integrate it into the story.

Gone are the days of scouring DeviantArt for images that only vaguely match the scene I envision. Now, I can create precisely what I need and refine it until it perfectly aligns with my vision.

The Wan 2.1 video LoRAs will be used for more risqué scenes I have planned. There’s a stark difference between reading about Rias and Akeno giving Issei a strip tease and seeing it unfold. By training these LoRAs, I can craft the backgrounds and characters exactly as I imagine them, bringing these moments to life in vivid detail.

I will be releasing these LoRAs publicly so that you, the audience, can download and use them locally. There are two main reasons for this. First, if you’re a prospective writer contributing to my thread, you’re welcome to download these LoRAs and use them to generate your own images of the characters. This ensures consistency in character appearances—everyone will have the same Rias, Akeno, Koneko, and so on, rather than multiple interpretations of the same character. Second, if you want to explore scenarios involving these characters that aren’t included in the story, you can use Wan 2.1 to create your own personalized scenes (wink, wink, nudge, nudge).

See below for the models.

Rias Gremory

Link to Wan 2.1 model: WIP

Link to Flux lora: https://civitai.com/models/1405984

Akeno Himejima

Link to Wan 2.1 model: WIP

Link to Flux lora: https://civitai.com/models/1471079/akeno-himejima-dxd