|
|
--- |
|
|
license: apache-2.0 |
|
|
base_model: |
|
|
- Wan-AI/Wan2.2-I2V-14B-480P |
|
|
tags: |
|
|
- lora |
|
|
- video-generation |
|
|
- image-to-video |
|
|
- pixel-art |
|
|
- sprite-animation |
|
|
- game-development |
|
|
- wan |
|
|
- comfyui |
|
|
library_name: diffusers |
|
|
pipeline_tag: image-to-video |
|
|
--- |
|
|
|
|
|
# Wan2.2 Pixel Animate Adapter |
|
|
|
|
|
A LoRA adapter for Wan 2.2 I2V (Image-to-Video) model, fine-tuned specifically for generating pixel art sprite animations from static images. |
|
|
|
|
|
## Model Details |
|
|
|
|
|
| Property | Value | |
|
|
|----------|-------| |
|
|
| Base Model | Wan2.2-I2V-A14B (14B parameters) | |
|
|
| Adapter Type | LoRA | |
|
|
| LoRA Rank | 256 | |
|
|
| Precision | bfloat16 | |
|
|
| File Size | ~2.3 GB | |
|
|
|
|
|
## Training |
|
|
|
|
|
- **Epochs**: 100 |
|
|
- **Optimizer**: AdamW (lr=2e-5, betas=[0.9, 0.99]) |
|
|
- **Gradient Accumulation**: 4 steps |
|
|
- **Activation Checkpointing**: Unsloth |
|
|
|
|
|
## Dataset |
|
|
|
|
|
Trained on **226 pixel art sprite animation videos** covering: |
|
|
|
|
|
- **Character Sprites**: Cowboys, zombies, skeletons, dark elves, fantasy characters, Santa, city characters, anime warriors |
|
|
- **Magic Effects**: Projectiles, elemental spells, energy bursts |
|
|
- **VFX**: Explosions, smoke effects, dust clouds |
|
|
- **Actions**: Attack cycles (slashing, shooting, casting), idle animations, walking cycles |
|
|
|
|
|
**Resolution**: 600x370 pixels |
|
|
**Frame Buckets**: 8, 16, 24, 32 frames (up to 2 seconds at 16fps) |
|
|
|
|
|
## Usage |
|
|
|
|
|
This LoRA is designed for **Image-to-Video generation** - transforming static pixel art characters into animated sprite sequences. |
|
|
|
|
|
### ComfyUI Workflow |
|
|
|
|
|
Load using `LoraLoaderModelOnly` node with the Wan 2.2 I2V model: |
|
|
|
|
|
1. Load base model: `wan2.2_i2v_high_noise_14B_fp16.safetensors` or `wan2.2_i2v_low_noise_14B_fp16.safetensors` |
|
|
2. Apply this LoRA adapter with strength 1.0 |
|
|
3. Use with `PainterI2V` node for image-to-video conditioning |
|
|
4. Recommended: Use with 4-step distillation LoRA for faster inference |
|
|
|
|
|
See included `wan2-2-video.json` workflow file for a complete setup. |
|
|
|
|
|
### Recommended Settings |
|
|
|
|
|
- **Sampler**: DDIM |
|
|
- **Steps**: 4 (with distillation LoRA) or higher without |
|
|
- **ModelSamplingSD3 Shift**: 5.0 |
|
|
- **Frame Count**: 45 frames |
|
|
- **CFG Scale**: 1.1 |
|
|
|
|
|
## Files |
|
|
|
|
|
- `wan2.2_animate_adapter_model.safetensors` - The LoRA adapter weights |
|
|
- `wan2-2-video.json` - ComfyUI workflow for using this model |
|
|
|
|
|
## License |
|
|
|
|
|
Please refer to the Wan 2.2 model license for usage terms. |
|
|
|
