SVI v2 Pro Installation Guide

// INITIALIZING SYSTEM: ComfyUI + Wan 2.2 + SVI Pro //

Phase 00

System & Inventory Check

Run these diagnostics in your ComfyUI_windows_portable folder to avoid redundant work.

// STEP 1: VERIFY TOOLS (PYTHON/GIT/GPU)python --version && git --version && nvidia-smi
// STEP 2: CHECK FOR EXISTING CUSTOM NODESdir custom_nodes /b | findstr /i "KJNodes WanVideoWrapper"

[ANALYSIS] If this outputs ComfyUI-KJNodes and ComfyUI-WanVideoWrapper, you can SKIP Phase 01.

// STEP 3: SCAN FOR LARGE MODEL FILESdir models\\diffusion_models /b | findstr /i "wan"

[ANALYSIS] If you see Wan2.2...safetensors files here, you can SKIP those downloads in Phase 02.

Phase 01

Install Custom Nodes

  1. Launch ComfyUI. Access the Manager interface.
  2. Select Install Custom Nodes.
  3. Locate and Install:
    • ComfyUI-KJNodes
    • ComfyUI-WanVideoWrapper
  4. REBOOT SYSTEM to apply changes.
Phase 02

Download Models

Transfer data to the designated directories. Precision is required.

Wan 2.2 High (FP8)

Kijai's Scaled Version (E4)
models/diffusion_models/
Download

Wan 2.2 Low (FP8)

Kijai's Scaled Version (E4)
models/diffusion_models/
Download

Light X2V (Accelerator)

High Noise LoRA (Speeds up 4x)
models/loras/Wan22_Lightx2v/
Download

SVI v2 Pro (High)

Infinity LoRA (Official)
models/loras/SVI/
Download

SVI v2 Pro (Low)

Infinity LoRA (Official)
models/loras/SVI/
Download

UMT5 XXL (FP8)

Text Encoder (Comfy-Org)
models/text_encoders/
Download

Wan 2.1 VAE

VAE Decoder (Comfy-Org)
models/vae/
Download
WARNING: You must create the subfolders models/loras/SVI and models/loras/Wan22_Lightx2v manually if they do not exist.
Phase 03

Initialize Workflow

Load the protocol file (JSON) to begin generation.

1. Access Official SVI GitHub.

2. Locate file: SVI_v2_Pro_Wan2.2.json.

3. Drag & Drop file into ComfyUI interface.

4. Execute Command: [QUEUE PROMPT].