javawock7618/comfy-wan2.2-workflows

Wan2.2 GGUF & INT8 Workflows Collection (with lightx2v LoRA)

This repository provides a collection of highly optimized ComfyUI workflows tailored for the Wan-AI/Wan2.2 model family using GGUF and INT8 quantized variants. Built around ComfyUI's native model loaders and quantized inference support, these workflows leverage lightx2v LoRA for accelerated generation, together with advanced features such as long video generation, character animation, motion transfer, and localized video inpainting and detail enhancement.

---

Workflow Files & Overview

The repository contains the following production-ready and experimental workflow files. (Note: YYMM.x in the file names represents the release year, month, and version number, e.g., 2606.1 or 2607.1)

#	File Name	Workflow Target	Key Features & Quick Notes
1	Wan2.2_SVI-javanoYYMM.x.json	Image-to-Video (SVI Long Gen)	Long-form video generation using Wan2.2 GGUF accelerated by lightx2v LoRA, with SVI2 Pro LoRA for high-fidelity i2v outputs.
2	Wan2.2_int8_SVI-javanoYYMM.x.json	Image-to-Video (SVI INT8 / Long Gen)	Highly optimized INT8 variant for SVI long video workflows. Provides improved inference throughput on supported GPUs compared with GGUF versions.
3	Wan_Animate2_int8-javanoYYMM.x.json	Wan Animate2 / Character Animation & Motion Transfer	INT8 workflow for Wan Animate2, designed for character animation, body-motion and facial-motion transfer, and character replacement from reference videos.
4	Wan2.2_Anime-javanoYYMM.x.json	Anime / Motion Transfer v2v	Video-to-Video workflow running on GGUF with lightx2v LoRA acceleration that extracts and transfers motion from reference videos.
5	Wan2.x_Inpaint-DetailerYYMM.x.json	Localized Inpaint & Detailer	An accelerated GGUF-driven v2v utility for partial correction and targeted regional enhancement beyond face-only detailers.

---

Detailed Workflow Breakdown

1 & 2. Image-to-Video / SVI Long Gen (SVI / SVI INT8)

Designed for robust, accelerated long-form video production utilizing Wan2.2's native Image-to-Video architecture.

• INT8 Computational Acceleration: The Wan2.2_int8_SVI-javanoYYMM.x.json workflow introduces INT8 support for SVI long video generation. On supported hardware, it can reduce inference time and improve throughput compared with standard GGUF pipelines.
• Dual LoRA Stacking (lightx2v + SVI2 Pro): Both SVI configurations are pre-tuned to stack the lightx2v LoRA for fast few-step sampling together with the SVI2 Pro LoRA for improved structural stability and detail retention.
• Extended Sampling Configuration: Customized temporal scheduling helps maintain visual consistency when generating videos beyond standard clip lengths.

3. Wan Animate2 / Character Animation (Animate2 INT8)

A specialized INT8 workflow based on Wan Animate2, designed for transferring character motion and facial expressions from reference videos. It can be used to animate a target character from a reference image or replace a character in a reference video while preserving the source performance.

• Character Motion Transfer: Transfers body movement, poses, gestures, and overall motion from a reference video to a target character.
• Facial Motion Transfer: Transfers facial expressions and related facial movements from the reference performance.
• Character Replacement: Replaces the original character in a reference video with a target character while maintaining the source performance.
• INT8 Inference: Uses an INT8-quantized model to reduce memory requirements and improve inference performance on supported hardware.
• Efficient Generation: The INT8 workflow is optimized for efficient inference within ComfyUI.

4. Anime / Motion Reference (Anime)

Specialized in traditional Video-to-Video (v2v) conversion utilizing fast GGUF inference and lightx2v acceleration.

• Motion Transfer Engine: Extracts structural motion from a source reference video and applies it to the generated subject.
• High-Speed Stylization Continuity: Leverages few-step sampling to transform reference motion into anime-style video while maintaining temporal consistency.

5. Localized Inpaint & Detailer (Inpaint-Detailer)

A versatile Video-to-Video utility designed for targeted corrections and regional enhancement.

• Universal Regional Inpainting: Isolate arbitrary regions such as faces, hands, clothing, background elements, or other detailed areas.
• Targeted Corrections: Rewrite or enhance selected areas while preserving the surrounding video.
• Accelerated Iteration: Uses lightx2v-based acceleration where supported to reduce rendering time during repeated correction passes.

---

Prerequisites & Installation

To ensure all workflows load correctly without errors, please ensure your environment is fully updated:

1. ComfyUI Core: Update ComfyUI to a recent version that supports the required native model loaders and quantized model formats.

2. Custom Nodes:

   • Ensure ComfyUI-GGUF or the corresponding native quantized model loader is available for GGUF workflows.
   • Ensure standard utility nodes such as ComfyUI-KJNodes are fully updated via ComfyUI Manager.
   • If any supporting node appears red upon loading, use ComfyUI Manager → Install Missing Custom Nodes.

3. Required Assets:

   • Base Models: Download the appropriate Wan2.2 GGUF or INT8 checkpoints and place them in the corresponding ComfyUI model directory.
   • Wan Animate2: The Wan_Animate2_int8-javanoYYMM.x.json workflow requires the corresponding Wan Animate2 INT8 model weights.
   • LoRAs: Place required LoRAs in your models/loras/ directory.
   • lightx2v LoRA: Required for workflows using lightx2v acceleration.
   • SVI2 Pro LoRA: Additionally required for the SVI and SVI INT8 workflows.