JoJo's Bizarre Adventure | style model

介绍：

1. 基于JOJO动画训练的风格模型（lycoris），它提供：

1. JOJO动画风上色
2. 替身使者风格的人物
3. 抽象化的背景

2. 版本区别

ver-n1：人物风格化程度适中，生成结果更偏向于人类。推荐权重 0.7～1
ver-s1：人物风格化程度较强，生成结果更偏向于人形替身使者。推荐权重 0.6～0.8

3. 推荐的SD模型及t2i参数

模型推荐：RevAnimated（最佳）、Dreamshaper、Lyriel、Anylora、AOM3、TmndMixP
当结果不理想时，可尝试调整lora的权重。
推荐图像宽高比为16:9，如960:540，根据需要的清晰度使用高清修复功能。

4. 提示词

触发词：jojos5standintro（有时也不是必须）
可选使用：stand(jojo)、jojo pose、style parody、jojo background、solo、cowboy shot
可以使用：stand(jojo)|xxx 语法来做融合，xxx可以是物体、特定人物、颜色。与分开成两个提示词相比，结果可能更多样化。

5. 图生图

使用Controlnet tile模型的固定工作流，可以非常稳定地将图像进行风格化（真人和二次元图像均可）。
denoising=0.8～1，CFG Scale=4～5，Steps=30～40
Controlnet preprocessor=tile_resample
Controlnet model: control_v11f1e_sd15_tile
Control Weight=1，starting step=0，ending step=1（至少0.4）
Control Mode=ControlNet is more important

————利用图生图或融合提示词，可能出现非常有创造性的生成结果。

*缺点

1. 生成图像大概率带有文字，且较难通过提示词排除。尝试使用其他图像处理工具或者局部重绘消除掉文字。
2. 生成人物的服装/设计种类不多，容易产生雷同感（这是因为只训练了第五部动画的替身使者）。可以通过增加服装、颜色、动作的提示词或使用提示融合语法来使结果多样化。

*未来计划

1. 替身使者更多样化。
2. 人物风格与替身使者风格分开调用。

祝您玩得愉快。

---

Introduction

1. A style lora model based on JOJO animation (lycoris). It provides:

1. JOJO animation coloring style
2. JOJO STAND style character
3. Abstracted colored background

2. Version description

ver-n1: The character stylization strength is moderate, and the result is more like a human. Recommended weight 0.71
ver-s1: The character stylization strength is strong, and the result is more like a STAND. Recommended weight 0.60.8

3. Recommended SD model and t2i params

Model recommendations: RevAnimated (best), Dreamshaper, Lyriel, Anylora, AOM3, TmndMixP
When the result is not good, try adjusting the lora weight.
Recommended image aspect ratio: 16:9 (e.g., 960:540), use hires fix based on desired clarity.

4. Prompt

Trigger: jojos5standintro (not always required)
Optional: stand(jojo), jojo pose, style parody, jojo background, solo, cowboy shot
Can use: stand(jojo)|xxx for prompt fusion. "xxx" can be an object, specific person, or color. Results may be more varied than using separate prompts.

5. Image to Image workflow

Using this workflow with the Controlnet tile model enables consistent stylization (works with both real-life and 2D/animation images).
For ver-s1: strength=0.8; for ver-n1: strength=1
denoising=1, CFG Scale=45, Steps=3040
Controlnet preprocessor: tile_resample
Controlnet model: control_v11f1e_sd15_tile
Control Weight=1, starting step=0, ending step=1 (minimum 0.4)
Control Mode: ControlNet is more important

———— Try using image2image with ControlNet or prompt fusion (a|b) for more creative results.

*To be improved

1. Generated images often contain text, which is hard to remove via negative prompts. Use image editing tools or inpainting to fix this.
2. Limited variety in costumes/designs for generated characters (due to training only on Season 5 STANDs). Enhance variety by adding more prompts for clothing, colors, or actions when using the lora.

*Future plans

1. Include more STANDS.
2. Allow separate use of character style and STAND style.

Have fun with it.