Perfect Video Resolution

A unified ComfyUI custom node that combines model-specific video resolution presets with smart image and mask resizing.

It merges the functionality of FindPerfectResolution and WanResolutions into a single node, making it easy to pick the right resolution for Wan 2.2, MiniMax H3, and LTX video generation while also resizing the input image(s) and mask to match.

Features

• Model selection: Wan 2.2, MiniMax H3, LTX
• Aspect-ratio driven resolution presets for each model
• Resolution quality tiers: Fast Draft, Preview, High Detail, Native (and model-specific labels)
• Optional image inputs for I2V/FL2V/SVI workflows:
  • image — main input image
  • image2 — secondary image, resized to the same resolution
  • mask — optional mask, resized with the same geometry
• Smart resize options: upscale, upscale method, small-image crop/pad, pad color
• Divisibility control: default 16 for Wan 2.2, 32 for MiniMax H3 / LTX
• Outputs width, height, resized IMAGE, IMAGE2, MASK, and a resolution_info string

Installation

Clone this repository into your ComfyUI custom_nodes folder:

cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/huchukato/ComfyUI-PerfectVideoResolution.git

Restart ComfyUI.

Usage

1. Add the node Perfect Video Resolution from the utils category.
2. Select the model, aspect_ratio, and resolution.
3. Optionally connect image, image2, and/or mask.
4. Connect width and height to your video sampler / latent generator.
5. Use the resized IMAGE, IMAGE2, and MASK outputs as needed.

When an image is connected, the node automatically picks the closest supported aspect ratio and resizes everything to the selected resolution.

Inputs

Name	Type	Description
model	COMBO	Video model: WAN 2.2, MiniMax H3, LTX
aspect_ratio	COMBO	Aspect ratio presets, depends on selected model
resolution	COMBO	Resolution quality tier, depends on model + aspect ratio
image	IMAGE	Optional input image (I2V/FL2V/SVI)
image2	IMAGE	Optional secondary image, resized to same resolution
mask	MASK	Optional mask, resized with same geometry
divisible_by	INT	Pixel alignment (default 16 for WAN, 32 for MMH3/LTX)
upscale	BOOLEAN	Allow upscaling when target is larger than source
upscale_method	COMBO	lanczos, bilinear, bicubic, nearest
small_image_mode	COMBO	none, crop, pad
pad_color	STRING	Background color for pad mode (default #000000)

Outputs

Name	Type	Description
width	INT	Calculated width
height	INT	Calculated height
IMAGE	IMAGE	Resized image
IMAGE2	IMAGE	Resized secondary image (placeholder if not connected)
MASK	MASK	Resized mask (placeholder if not connected)
resolution_info	STRING	Human readable resolution summary

Credits

Built by combining ideas from:

• comfyui-find-perfect-resolution
• ComfyUI-WanResolutions