本模型采用《龙珠Z》动画风格，由676张精选截图训练而成，具有16维度的模型精度，画风匹配度高，且具有良好的泛化能力，以此向鸟山明老师致敬。
This model is styled after ‘Dragon Ball Z’ animation, trained on 676 selected screenshots, with a model dimension of 16, achieving high style fidelity and good generalization, as a tribute to Toriyama Akira sensei.
Description
1. 使用Flux dev2pro作为基础模型，以降低图像崩溃的风险。
2. 增加dropout比率，以获得更好的泛化能力。
3. 使用t5_attn_mask技术。
1. Using Flux dev2pro as the base model to reduce the probability of image collapse.
2. Increasing the dropout rate for better generalization.
3. Employing the t5_attn_mask technique.
Details
Downloads
157
Platform
CivitAI
Platform Status
Available
Created
10/5/2024
Updated
9/28/2025
Deleted
-
Trigger Words:
DragonBall_Z_style
Files
DragonZ_v4.3_16dim.safetensors
Mirrors
CivitAI (1 mirrors)