Yamashita Mitsuki - Wan2.2 T2V LoRA (High-Noise)

High-Noise Character LoRA for Wan 2.2 Text-to-Video model.

Available Checkpoints

This repository contains 10 checkpoints (Epoch 3, 6, 9, 12, 15, 18, 21, 24, 27, 30).

Recommended: Use yamashita_mitsuki_HIGH.safetensors (Final version, Epoch 30)

Model Details

  • Type: High-Noise LoRA (Timestep 875-1000)
  • Base Model: Wan 2.2 T2V 14B (High-Noise)
  • Training Dataset: 30 images
  • Training Epochs: 30 (7,200 steps)
  • Network Dim/Alpha: 32/32
  • Learning Rate: 1e-4
  • Final Loss: 0.0152
  • Resolution: 768x768

Usage

Important: This LoRA must be used together with Low-Noise LoRA

Required Files:

Settings:

  • LoRA Weight: 0.8-1.0
  • Use both HIGH and LOW together
  • Compatible with ComfyUI

Training Details

  • Optimizer: AdamW8bit
  • LR Scheduler: cosine_with_min_lr
  • FP8: Enabled
  • Discrete Flow Shift: 5.0
  • Training Time: ~5 hours on A100 80GB

Citation

Trained with musubi-tuner by kohya-ss.

Downloads last month
-
Inference Providers NEW

This task can take several minutes

Model tree for Shuntea/yamashita_mitsuki_wan22_HIGH

Adapter
(52)
this model