Skip to content

WAN22_NODES_README.md

Latest commit

 

History

History
363 lines (269 loc) · 8.91 KB

WAN22_NODES_README.md

File metadata and controls

363 lines (269 loc) · 8.91 KB
Raw

Wan 2.2 Video Generation Nodes

Complete toolkit for generating optimized dimensions for Wan 2.2 video generation in ComfyUI.

Overview

Two complementary nodes that generate Wan 2.2 compliant video dimensions:

Node Best For Input Ratios
Aspect Ratio Helper I2V workflows IMAGE (required) Any (auto-detected)
Size Preset T2V workflows None (dropdowns only) 15 predefined

Both nodes follow official Wan 2.2 specifications:

  • ✅ Divisibility by 8 pixels
  • ✅ Aspect ratio range: 1:3 to 3:1 (0.333 to 3.0)
  • ✅ Smart hybrid algorithm using 28 documented training sizes

Quick Start

For Image-to-Video (I2V)

Load Image → Wan 2.2 Aspect Ratio Helper → Empty Latent Video
- Auto-detects aspect ratio from input image
- Generates optimized dimensions matching your reference

For Text-to-Video (T2V)

Wan 2.2 Size Preset → Empty Latent Video
- Select aspect ratio from 15 presets
- Choose size preset (tiny to gigantic)
- Perfect for starting from scratch

Wan 2.2 Aspect Ratio Helper v2.2

Features

  • 📷 Analyzes input image aspect ratio automatically
  • 🎯 Generates 6 optimized sizes (tiny → gigantic)
  • 🔄 Change size preset without rewiring
  • 📊 Smart hybrid: uses known Wan 2.2 sizes when available
  • ⚡ Handles any ratio from 1:3 to 3:1

Inputs

  • image: IMAGE tensor (required)
  • size_preset: tiny | small | medium | large | extra-large | gigantic (default: medium)

Outputs

  • width: INT - Selected width based on preset
  • height: INT - Selected height based on preset
  • info_text: STRING - Formatted display of all 6 sizes

Example

Load Image (1920×1080)
    ↓
Aspect Ratio Helper
├─ Detects: 16:9 ratio (1.778)
├─ size_preset: "medium"
└─ Output: 1080×608

Wan 2.2 Size Preset v1.1

Features

  • 📐 15 predefined aspect ratios
  • 🎯 6 size presets (tiny → gigantic)
  • 🚫 No image input needed
  • 🔄 Change ratio or size without rewiring
  • ⚡ Perfect for platform-specific formats

Inputs

  • aspect_ratio: 15 predefined options (default: "16:9 Widescreen")
  • size_preset: tiny | small | medium | large | extra-large | gigantic (default: medium)

Outputs

  • width: INT - Selected width
  • height: INT - Selected height
  • info_text: STRING - Formatted display of all 6 sizes

Available Aspect Ratios

Portrait (Vertical)

  • 1:3 Ultra-Tall Portrait
  • 9:21 Tall Portrait
  • 9:19.5 Tall Portrait
  • 9:16 Tall Portrait (Instagram, TikTok)
  • 3:4 Portrait
  • 5:6 Portrait

Square

  • 1:1 Square (Instagram)

Landscape (Horizontal)

  • 6:5 Landscape
  • 4:3 Standard
  • 16:10 Widescreen
  • 16:9 Widescreen (YouTube, HD)
  • 19.5:9 Widescreen
  • 21:9 Ultra-Wide (Cinematic)
  • 3:1 Ultra-Wide Landscape

Example

Size Preset
├─ aspect_ratio: "16:9 Widescreen"
├─ size_preset: "medium"
└─ Output: 1080×608

Size Presets

Both nodes offer 6 size options:

Preset Target Pixels Example (16:9) Use Case
Tiny ⚡ ~200K 592×336 Ultra-fast testing, iteration
Small ~400K 848×480 Quick tests
Medium ✨ ~650K 1080×608 Standard quality (default)
Large ~900K 1280×720 High quality
Extra-Large ~1.4M 1600×904 Very high quality
Gigantic 🎯 ~2M 1896×1072 Maximum quality, production

Recommended Workflow

1. TINY → Test ideas quickly (seconds)
2. SMALL/MEDIUM → Refine composition
3. LARGE → Quality validation
4. GIGANTIC → Final production render

Platform Quick Reference

YouTube (16:9 Widescreen)

Size Preset
├─ aspect_ratio: "16:9 Widescreen"
├─ size_preset: "large"
└─ Output: 1280×720

Instagram Reel / TikTok (9:16 Portrait)

Size Preset
├─ aspect_ratio: "9:16 Tall Portrait"
├─ size_preset: "medium"
└─ Output: 408×720

Instagram Square (1:1)

Size Preset
├─ aspect_ratio: "1:1 Square"
├─ size_preset: "medium"
└─ Output: 720×720

Cinematic (21:9 Ultra-Wide)

Size Preset
├─ aspect_ratio: "21:9 Ultra-Wide"
├─ size_preset: "extra-large"
└─ Output: 1408×608

Decision Matrix

Use this to pick the right node:

Your Situation Use This Node
Have reference image Aspect Ratio Helper
Starting from scratch Size Preset
I2V workflow Aspect Ratio Helper
T2V workflow Size Preset
Need exact ratio from artwork Aspect Ratio Helper
Know target platform Size Preset
Custom/unusual ratio Aspect Ratio Helper
Standard ratio (16:9, 9:16) Size Preset

Technical Details

Smart Hybrid Algorithm

Both nodes use the same intelligent sizing:

  1. Check Known Sizes: Searches 28 documented Wan 2.2 training sizes
  2. Match Criteria: Ratio within 5% AND pixels within 30%
  3. Use Documented: If match found, use known size (better model performance)
  4. Calculate Optimal: If no match, calculate ideal dimensions
  5. Round to 8: Ensures all dimensions divisible by 8

Known Wan 2.2 Training Sizes (28 total)

Square: 720×720

Portrait: 560×720, 560×896, 560×1088, 560×1280, 688×880, 688×1072, 848×1088

Landscape: 720×560, 896×560, 1088×560, 1280×560, 880×688, 1072×688, 1088×848, 1280×720, 1280×800, 1280×544, 1600×640

Additional: 1088×672, 672×1088, 1088×704, 704×1088, 1072×672, 672×1072, 1280×768, 768×1280

Size Calculation Formula

When no known size matches:

height = sqrt(target_pixels / aspect_ratio)
width = aspect_ratio × height
# Round both to nearest multiple of 8

Integration

Both nodes integrate seamlessly with:

Multi-LoRA Loaders

[Aspect Ratio Helper OR Size Preset]
├─ width ──→ Empty Latent Video (width)
└─ height ──→ Empty Latent Video (height)

Multi-LoRA Loader (Wan T2V/I2V/Qwen)
└─ model ──→ Wan Model Node

Standard ComfyUI Nodes

  • Empty Latent Video
  • Show Text (for info_text)
  • Note (for documentation)
  • Reroute (for clean connections)

Console Output

Both nodes print detailed information:

==================================================
Wan 2.2 Aspect Ratio Helper
==================================================
Input Image Dimensions: 1920×1080
Detected Aspect Ratio: 16:9 Widescreen (1.778)

All Available Sizes:
   tiny: 592×336 (198,912 pixels)
   small: 848×480 (407,040 pixels)
 → medium: 1080×608 (656,640 pixels)
   large: 1280×720 (921,600 pixels)
   extra-large: 1600×904 (1,446,400 pixels)
   gigantic: 1896×1072 (2,032,512 pixels)

Selected Output: 1080×608
==================================================

Troubleshooting

Q: Node not appearing in menu?

A: Restart ComfyUI - new nodes load on startup.

Q: Dimensions seem wrong?

A: Check console output for detailed breakdown. All dimensions guaranteed divisible by 8.

Q: Which size preset to use?

A: Start with "medium" (default) for best balance. Use "tiny" for testing, "gigantic" for final renders.

Q: Can I add custom ratios to Size Preset?

A: Yes! Edit ASPECT_RATIOS dictionary in Wan22_Size_Preset.py.

Q: Input image ratio out of range?

A: Aspect Ratio Helper automatically clamps to valid range (1:3 to 3:1) with console warning.

Version History

v2.2 / v1.1 (2025-10-24)

  • ✨ Added "tiny" preset (~200K pixels) for ultra-fast testing
  • ✨ Added "gigantic" preset (~2M pixels) for maximum quality
  • Now 6 size presets total

v2.1 / v1.0 (2025-01-23)

  • ✨ Simplified to 3 outputs (width, height, info_text)
  • ✨ Size Preset node created as T2V companion
  • ✨ Change presets without rewiring

v2.0 (2025-01-23)

  • ✨ Dynamic calculation with official Wan 2.2 specs
  • ✨ Divisibility by 8 (not 16)
  • ✨ Full ratio range 1:3 to 3:1

v1.0 (2025-01-23)

  • 🎉 Initial Aspect Ratio Helper release

Files

Node Files

  • nodes/Wan22_AspectRatio_Helper.py - Image-based aspect ratio node
  • nodes/Wan22_Size_Preset.py - Preset-based size selection node

Documentation

  • documentation/WAN22_NODES_README.md - This file
  • CHANGELOG.md - Detailed version history

Credits

Author: Eric
Category: Wan 2.2 / Video Generation
License: MIT
Created: January 2025
Updated: October 2025

Related Resources

  • ComfyUI - Main framework
  • Wan 2.2 Model - Video generation model
  • Multi-LoRA Loaders - Compatible LoRA management nodes

Support

For issues, feature requests, or questions:

  1. Check console output for error details
  2. Review this documentation
  3. Open an issue on GitHub

Quick Links: