Nourepide-ComfyUI-Allor

Allure is a plugin for ComfyUI with an emphasis on transparency and performance.
All modules support transparency, multi-image and are easy to use.
One of the main features is the ability to combine several images of different sizes.
Almost all functionality is implemented in tensor space without transformation into an image, where possible. This allows you to achieve an incredible speed of work, in some places the speed increases from 1-2 seconds to 4-6 milliseconds.

Installation:

cd custom_nodes
git clone https://github.com/Nourepide/comfyui_allor.git comfyui_allor
cd comfyui_allor
pip install -r requirements.txt
You can also create fonts folder in comfy_extras or just start ComfyUI and folder will create automatically.

Modules:

Details:

Alpha chanel

Helpers for work with alpha chanel in exiting images.

Nodes:

By Mask

Applies transparency to the image using a mask.

Params:

method -
- default - Standard mask overlay.
- invert - Inverted mask overlay.

Restore

Add alpha dimension to images. Or restore if it already exists.

Params:

method -
- default - Apply adding new alpha chanel or restore chanel values to max if it already exists.
- only_add - Only add new alpha chanel if it does not exist.
- only_restore - Only restore alpha chanel values to max if it already exists.

Image Container

Creating blank images of fixed or inherited size.

Nodes:

All

[red, green, blue] [0 - 255] - Color of the container in RGB.
alpha [0.0 - 1.0] - Transparency of the container.

Default

Create an empty container.

Params:

[width, height] [0 - *] - Size of a new container.

Inheritance Add

Get images size and create a container with adding value to this size.

Params:

add_[width, height] [0 - *] - The value that will then be added to the size of the container.

Inheritance Scale

Get images size and create a container with scaling this size.

Params:

scale_[width, height] [0.0 - *] - Scale factor that will then be change the size of the container.

Inheritance Max

Get max width and max height from images_a and images_b and create a new container with this size.

Inheritance Sum

Get width and/or height from images_a and images_b and create a new container with sum this size.

PS:

The first letter i in the word Inheritance is barely noticeable. But in the end, I decided to leave it as it is. At first, I thought to replace this word with the word Extended, but one would think that this node expands the incoming images by adding size to it, and does not create a new image inheriting size from the incoming images.

Image Composite

Allows you to place one image on top of another.

Nodes:

All

background - The placement order of images in the container.
method - The method of connecting images makes sense only when 2 or more images are passed to the input.

Methods:

Inputs - [a1, a2, a3] and [b1, b2, b3].
Outputs -
    pair: [a1, b1], [a2, b2], [a3, b3]
    matrix: [a1, b1], [a1, b2], [a1, b3]
            [a2, b1], [a2, b2], [a2, b3]
            [a3, b1], [a3, b2], [a3, b3]

Absolute

Use it for pixel-perfect images compositing.

Params:

images_[a, b]_x [0 - *] - Images horizontal position. [left -> right].
images_[a, b]_y [0 - *] - Images vertical position. [top -> bottom].
container_[width, height] [0 - *] - Container size. If 0, then the container will be with the maximum width and height of the images.

Relative

Use it for percentage images compositing.

Params:

images_[a, b]_x [0.0 - 1.0] - First images horizontal offset. [left edge -> right edge].
images_[a, b]_y [0.0 - 1.0] - First images vertical offset. [top edge -> bottom edge].
container_size_type
- max - Maximum of width and height from images_a and images_b.
- sum - Sum of width and height from images_a and images_b.
- sum_[width, height] -

How it works.

Images in relative mode cannot go beyond the edges of the container.
In horizontal [0.0 - images is located on the left of the container], [1.0 - on right of container].
In vertical [0.0 - images is located on the top of the container], [1.0 - on bottom of container].
In all cases, [0.5] means that the images will be in the center.

Image Segmentation

Warning!: Required ONNX. ONNX currently doesn't support Python 3.11 and above. See details.

Nodes:

All

alpha_matting [boolean] - Alpha metting is a post-processing step that can be used to improve the quality of the output.
alpha_matting_foreground_threshold [0 - 250] - Trimap foreground threshold.
alpha_matting_background_threshold [0 - 250] - Trimap background threshold.
alpha_matting_erode_size [0 - *] - How far will the transparency spread from the original mask.
post_process_mask [boolean] - Post-processing mask, makes it rougher.

Default

Allows you to select a model with pre-specified settings.:

Params:

u2net - GitHub
u2netp - GitHub
u2net_human_seg - GitHub
u2net_cloth_seg - GitHub
silueta - GitHub
isnet-general-use - GitHub
isnetis - GitHub
modnet-p - GitHub
modnet-w - GitHub

Custom

Allows you to select a model from the onnx folder and specify the parameters yourself.

Params:

model [*.onnx] - Your model in ComfyUI/models/onnx folder.
mean [0.0 - 1.0] - Clarifying value.
std [0.0 - 1.0] - The divisor of the final value. I recommend leaving it at 1.0.
size [0 - *] - Title size. Different models support different sizes. Or their multiple value [512 -> 1024 -> 2048...]. If the size does not match the desired one, you will see a notification in the console.

Custom Advanced

Allows you to select a model from the onnx folder and finely specify the parameters yourself.

Params:

model [*.onnx] - Your model in ComfyUI/models/onnx folder.
mean_r [0.0 - 1.0] - Clarifying value. Red chanel.
mean_g [0.0 - 1.0] - Clarifying value. Green chanel.
mean_b [0.0 - 1.0] - Clarifying value. Blue chanel.
std_r [0.0 - 1.0] - The divisor of the final value. Red chanel.
std_g [0.0 - 1.0] - The divisor of the final value. Green chanel.
std_b [0.0 - 1.0] - The divisor of the final value. Blue chanel.
width [0 - *] - Width of title size.
height [0 - *] - Height of title size.

Some models support non-standard title size. For example [w:1024 - h:2048].

PS:

Made on rembg. The solution to some problems may be dependent on the rembg developer.

ImageText

Creating text as image.

Fonts should be contained in ComfyUI/comfy_extras/fonts.

You also can change the fonts folder in config.

Nodes:

Default

Create an image with text.

Params:

text - Your input. Shouldn't be empty.
font [*.otf, *.ttf] - Font from ComfyUI/comfy_extras/fonts folder.
size [0 - *] - Font size in SP. Warning!: It's not PX size of the container. The size of the container is calculated using a variety of parameters. Example: o and j will have different height and exactly different size compared to 树.
[red, green, blue] [0 - 255] - Color of the text in RGB.
alpha [0.0 - 1.0] - Transparency of the text.
margin_x [0 - *] - Horizontal offset.
margin_y [0 - *] - Vertical offset.

Outlined

Create an image with outlined text.

Params:

outline_size [0 - *] - Outline size in PX.
outline_[red, green, blue] [0 - 255] - Color of the outline in RGB.

Clamp

Clamps help to control the "web" of nodes and improve readability.

Nodes:

ClipClamp
ClipVisionClamp
ClipVisionOutputClamp
ConditioningClamp
ControlNetClamp
GligenClamp
ImageClamp
LatentClamp
MaskClamp
ModelClamp
StyleModelClamp
UpscaleModelClamp
VaeClamp

Configuration:

You can configurate Allore with config.json.

fonts_folder_path - Array with a relative path to fonts folder, by default is ["comfy_extras", "fonts"]. (Converted to ComfyUI/comfy_extras/fonts for Unix and ComfyUI\comfy_extras\fonts for Windows).
modules - You can disable (or enable) modules at will. Nodes from disabled modules will not be loaded during the start of ComfyUI.