Wan 2.1: control-loras

Introduction

This workflow is based on the Control-LoRAs low-rank adaptation technology of the WAN2.1 framework and provides two core functions:

1. Depth control: Generate depth maps through AI-driven 3D scene analysis to accurately construct the three-dimensional spatial structure and perspective relationship of architectural visualization and game scenes;

2. Tile intelligent block: Use video memory optimization strategy to achieve image block generation and detail recursive enhancement, support 4K/8K ultra-high-definition rendering and texture repair of old materials.

https://huggingface.co/spacepxl/Wan2.1-control-loras

https://huggingface.co/Kijai/WanVideo_comfy/tree/main

Workflow Overview

How to use this workflow:

Part 1 ：depth

Step 1: Select workflow 'depth'

Step 2: Upload and zoom videos

1.Upload Video.

2.Select the resolution and scaling method for the video（If the video is not scaled, it may cause an error）.

Step 3: Select depth lora model

Step 4: Adjust the start and end positions of the control

Step 5: Adjust the frame rate of video output

Part 2：tile

Step 1: Select workflow 'tile'

Step 2: Upload and zoom videos

1.Upload Video.

2.Select the resolution and scaling method for the video（If the video is not scaled, it may cause an error）.

Step 3: Select tile lora model

Step 4: Adjust the start and end positions of the control

Step 5: Adjust the frame rate of video output