ArtificialGuyBR

Home / Blog / Training LoRAs

Wan 2.2 Disintegration LoRA: Turn Images into Particle Effects

LoRA for Wan 2.2 I2V that creates dramatic disintegration particle effects from static images using the trigger word 'Disintegration'.

2 sources cited Training LoRAs

On this page

  1. What This Model Does
  2. Trigger Word & Usage
  3. Files & Installation
  4. ComfyUI Workflow
  5. Recommended Settings
  6. Common Errors and Fixes
  7. FAQ
  8. Sources

Wan 2.2 Disintegration LoRA: Turn Images into Particle Effects

TL;DR

What This Model Does

The Disintegration Redmond LoRA adds a specialized disintegration effect to Wan 2.2 Image-to-Video generation. Feed it a static image and a prompt containing the word "Disintegration", and it animates the subject breaking apart into particles, dust, or energy that dissolves into the air. The effect suits magical transformations, sci-fi teleportation sequences, supernatural vanishings, and dramatic scene transitions.

The LoRA was trained on the Wan-AI/Wan2.2-I2V-A14B base model (14B parameters) and released under CC0-1.0, meaning you can use it commercially without attribution. Redmond AI sponsored the training compute.

Trigger Word & Usage

Trigger word: Disintegration

Include this exact word in your prompt when describing the scene. The model recognizes it and applies the disintegration animation to your input image.

Example prompts from the model card:

(Keep prompts simple; the trigger word does the heavy lifting.)

Files & Installation

Two SafeTensor files are provided:

Each is ~73 MB. Download both and place them in:

ComfyUI/models/loras/

ComfyUI Workflow

  1. Load your base Wan I2V model (Wan2.2-I2V-A14B)
  2. Add a Load LoRA node after the model loader
  3. Connect the model loader output to the Load LoRA model input
  4. Select both LoRA files in the Load LoRA node
  5. Connect the Load LoRA MODEL output to your KSampler
  6. Set strength_model to 0.8–1.0 (start at 1.0, reduce if artifacts appear)

Setting | Value

|---|---

Base Model | Wan-AI/Wan2.2-I2V-A14B LoRA Strength | 0.8–1.0 Trigger Word | Disintegration Pipeline | diffusers / ComfyUI License | CC0-1.0

Common Errors and Fixes

Error | Cause | Fix | Source

|---|---|---|---

No disintegration effect | Trigger word missing or misspelled | Ensure Disintegration appears exactly in prompt | Model card Artifacts / overcooked video | LoRA strength too high | Reduce strength_model to 0.8–0.9 | Model card Model not loading | Wrong base model | Use Wan2.2-I2V-A14B, not T2V or other variants | Model card Only one LoRA applied | Only one file selected | Select both low_noise and high_noise in Load LoRA | Model card

FAQ

Q: Does this work with Wan T2V (Text-to-Video)? A: No. This LoRA is trained specifically for Wan 2.2 I2V (Image-to-Video). It expects an input image.

Q: Can I use just one of the two LoRA files? A: The model card instructs to use both low_noise and high_noise together for the intended effect.

Q: Is this free for commercial use? A: Yes, CC0-1.0 license permits commercial use without attribution.

Q: What resolution / aspect ratios work best? A: The model card shows example outputs at 1168×776 and 664×1000. Wan 2.2 I2V natively supports multiple resolutions; start with the base model's recommended sizes.

Sources