🚀 Wan 2.2 & FLUX Krea Full Tutorial — Automated Install & Perfect Presets
This repository contains the full guide and resources for installing Wan 2.2 and FLUX Krea with a 1-click automated installer. Utilize our pre-made, optimized presets—developed after hundreds of parameter tests—to generate the highest quality videos and images. This tutorial leverages SwarmUI with a ComfyUI backend to make advanced AI content creation accessible, even on lower-end hardware.
🎬 Watch the Full Video Tutorial
Click the thumbnail below to watch the complete step-by-step video guide on YouTube.
Direct Link: https://youtu.be/8MvvuX4YPeo
✨ Key Features of This Setup
- Automated 1-Click Installation: Get Wan 2.2, FLUX Krea, SwarmUI, and ComfyUI running with minimal effort.
- Optimized Presets: Access a pack of ready-to-use presets for generating top-tier videos and images without manual tweaking.
- FLUX Krea DEV Model: Learn to generate significantly better images with the new FLUX Krea DEV model compared to the standard FLUX Dev.
- Low-End Hardware Friendly: The SwarmUI & ComfyUI combination is optimized to run on consumer-grade GPUs.
- Comprehensive Guidance: From downloading prerequisites to generating your first masterpiece, every step is covered.
🛠️ Essential Resources & Links
💾 Core Downloads
- SwarmUI Installer & AI Models Downloader: Contains the SwarmUI installer, AI model downloader app, presets, and prompt generator guide.
- Advanced ComfyUI 1-Click Installer: Includes support for Flash Attention, Sage Attention, xFormers, Triton, DeepSpeed, and RTX 5000 series cards.
🎓 Related Tutorials
- Main SwarmUI Installation Guide: Watch on YouTube
- RunPod Cloud Install Guide: Watch on YouTube
- Massed Compute Cloud Install Guide: Watch on YouTube
- Prerequisites (Python, Git, CUDA, FFMPEG, etc.): Watch on YouTube
💬 Community & Support
- Official SECourses Discord (10,500+ Members): Join the Community
- Stable Diffusion & AI Resources on GitHub: Explore on GitHub
- Official SECourses Reddit: Subscribe on Reddit
⚙️ Presets Overview
Our custom presets are designed to give you the best results out of the box. Import them into SwarmUI and start creating immediately.
📖 In-Depth: Exploring Wan 2.2
Wan 2.2 represents a significant leap forward in AI-driven video and image generation, focusing on cinematic quality, efficiency, and user control.
What is Wan 2.2?
Wan 2.2 is a generative AI model optimized for the "language of film." It excels in text-to-video (T2V) and image-to-video (I2V) generation, producing realistic, high-resolution content from descriptive prompts. It supports resolutions up to 720p at 24fps and can generate clips up to 5 seconds long. As an open model, it's accessible via platforms like Hugging Face and integrated into tools like ComfyUI.
Key Technical Features
- Mixture-of-Experts (MoE) Architecture: Separates the denoising process across different timesteps using specialized expert models, expanding capacity while maintaining computational efficiency.
- Cinematic-Level Aesthetics: Trained on curated data with detailed annotations for lighting, composition, and color, enabling precise control over cinematic styles.
- Enhanced Image and Video Generation: Improved detail in textures, higher-resolution rendering, and precise style controls.
- Advanced Camera Control: Mathematical precision for camera movements like dolly, pan, and crane shots, including motion blur and speed ramping.
- Temporal Consistency: Ensures coherent frame-to-frame transitions, maintaining character persistence and smooth motion.
- SNR Switching Mechanism: Switches between high-noise (structure-building) and low-noise (detail-refining) experts during generation, mimicking a professional workflow.
Model Variants
Wan 2.2 is available in multiple variants for different use cases:
| Model Variant | Parameters (Active) | Capabilities | Key Files |
|---|---|---|---|
| TI2V-5B | 5 Billion | Text-to-Video & Image-to-Video | wan2.2_ti2v_5B_fp16.safetensors, wan2.2_vae.safetensors |
| T2V-A14B | 14 Billion (MoE) | Text-to-Video | wan2.2_t2v_high_noise_14B_fp8_scaled.safetensors, wan2.2_t2v_low_noise_14B_fp8_scaled.safetensors, umt5_xxl_fp8_e4m3fn_scaled.safetensors, wan_2.1_vae.safetensors |
| I2V-A14B | 14 Billion (MoE) | Image-to-Video | wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors, wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors, umt5_xxl_fp8_e4m3fn_scaled.safetensors, wan_2.1_vae.safetensors |
System Requirements & Integration
- Hardware: The
TI2V-5Bmodel runs on consumer GPUs like an RTX 4090 (24GB VRAM recommended). TheA14Bmodels require multi-GPU setups (8x GPUs) or cloud services for optimal performance. - Integration: Easily integrated via ComfyUI visual workflows, Diffusers, and APIs.
Use Cases
- Digital Art & Illustration: Generating concept art or animated illustrations.
- Advertising: Creating promotional videos and product visuals.
- Film & Pre-Visualization: Storyboard animations and scene prototypes.
- Game Development: Designing characters, environments, and skill effects.
- Social Media: Quick, high-quality short-form video content.
🕒 Video Chapters
Navigate to the exact section you need with these timestamps.
- 0:00 Introduction: The Ultimate Wan 2.2 Tutorial
- 1:03 Free Prompt Generation Tool & Introducing FLUX Krea Dev
- 2:01 How SwarmUI & ComfyUI Enable Generation on Low-End Hardware
- 2:46 Quick Start: Downloading the Installers
- 3:10 Step-by-Step: Fresh Installation of ComfyUI
- 3:51 Step-by-Step: Fresh Installation of SwarmUI
- 4:18 Configuring the SwarmUI Backend for ComfyUI
- 4:53 One-Click Setup: Downloading All Required Models
- 5:46 Importing the Ultimate SwarmUI Presets Pack
- 6:22 Wan 2.2 Image-to-Video Generation Guide
- 7:33 How to Generate Amazing Prompts for Free
- 8:12 Starting Your First Generation & Monitoring Logs
- 8:53 Pro Tip: How to Fix Low GPU Utilization for Max Speed
- 10:32 Wan 2.2 Text-to-Video: Choosing the Right Preset
- 11:22 Generating a Detailed Dinosaur Animation Scene
- 12:15 In-Depth Analysis: 8 Steps vs 20 Steps & LoRA Impact
- 13:11 Finding the Best Parameters: CFG Scale & Step Counts
- 13:42 Advanced Optimization: Using TeaCache
- 15:31 FLUX Krea Dev vs FLUX Dev: Side-by-Side Comparison
- 16:26 How to Easily Train Your Own LoRAs on FLUX Krea Dev
- 17:02 Complete Workflow for High-Quality Images with FLUX Krea Dev
- 18:20 The Final Verdict: Side-by-Side Result Comparison
- 19:20 Experiment: Generating Still Images with Wan 2.2
- 21:18 Final Thoughts, Summary, and What’s Coming Next
