ComfyUI Extension: ComfyUI_RaykoStudio

Authored by Raykosan

Created about a year ago

Updated 5 days ago

72 stars

Run ComfyUI workflows without the setup

No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

A node for ComfyUI that allows you to overlay text on an image in an area defined by a mask, with support for text rotation, custom fonts, line breaks and transparency.

README

🦊 ComfyUI_RaykoStudio

Set of custom nodes for ComfyUI providing additional image processing capabilities

Nodes do not require the installation of additional Python packages.
🖥️ Performance has been tested for:

ComfyUI 0.15 and above
Python 3.10 and above
Torch 2.8 and above
Cuda 12.4 and above

📃 Requirements

torch
opencv-python
numpy
Pillow>=10.0.0
freetype-py>=2.5.1
pycairo>=1.29.0
scipy>=1.10.0

🛠 Installation

Set of nodes can be installed in several ways:

Clone repository to ComfyUI/custom_nodes/ folder:

git clone https://github.com/Raykosan/ComfyUI_RaykoStudio.git

Copy ComfyUI_RaykoStudio folder to: ComfyUI/custom_nodes/
You can install this node using the ComfyUI_Manager
To update ComfyUI_RaykoStudio, use the update.bat file

🦊 RS Collage Node

Interactive node for overlaying images with real-time positioning, scaling, rotation, and edge feathering directly on the canvas.

https://youtu.be/nPSujGTI_7s

Normal Mode
<img width="446" height="708" alt="Screenshot_1" src="https://github.com/user-attachments/assets/cd7da099-511a-4f28-a71e-54612f0c7f30" />

Advanced Mode
<img width="2106" height="955" alt="Screenshot_2" src="https://github.com/user-attachments/assets/0104c418-9126-4c11-8f2d-67f0969edb48" />

🔥 Features

Interactive Canvas — Drag, resize, and rotate overlays with visual handles
Center-Anchored Scaling — Corner handles scale proportionally; edge handles scale along a single axis, both expanding/contracting from the geometric center
Real-Time Feathering Preview — Supports Radial Blur In, Radial Blur Out, Ellipse Blur In and Ellipse Blur Out modes with adjustable radius
Non-Blocking Session — Waits indefinitely for ✔️APPLAY or ❌CANCEL input without hard timeouts or queue interruption
Precise Coordinate Mapping — Maintains a frozen viewport matrix during interaction to prevent drift; converts relative normalized coordinates to absolute pixel transforms for the backend
Integrated Controls — Opacity slider, flip toggles, and feather parameters accessible via node widgets
Workflow is in the Example folder — You can create an endless chain of these nodes - you connect the image output to the background input of the next node

🖼️ Working Modes

Normal Mode
Quick editing directly inside the ComfyUI node. Perfect for simple tasks and final touches.

Advanced Mode
Fullscreen editor with side panel for fine-tuning:

Real-time display across the entire screen
Independent zoom (mouse wheel) and canvas panning
Camera automatically fits composition to 80% of window
Overlay can extend beyond background boundaries without scale recalculation
All parameters adjustable via convenient sliders

🔆 Feathering Types

| Type | Description | |------|-------------| |None|No feathering| |Radial Blur In|Radial blur from center to edges| |Radial Blur Out|Inverted radial blur| |Ellipse Blur In|Elliptical blur from center| |Ellipse Blur Out|Inverted elliptical blur|

💡 Each feathering type allows you to drag the effect center directly on the overlay.

🪛 Usage

For the overlay, it is better to use images with transparency (PNG, WebP, and TIFF files containing transparency) or images coming from the background removal node (RMBG). If you want to use a regular image or an RGB image with transparency for the overlay, but not RGBA, upload it via the 🦊 RS rgb2rgba node. Node will make or correct the Alpha channel correctly.
Connect the tensors background_image and overlay_image to the node and start the generation.
Adjust the overlay using the markers on the canvas:

Corners - Proportional scaling from the center
Edges - Scaling on one axis from the center
Red cross - Freely movable blur center
Top yellow marker - Rotate around the center

Adjust the type of shading, radius, and opacity using widgets. After selecting the type of shading, a red cross will appear on the overlay, which can be used to specify the center of the shading or blur.
Click ✔️APPLY to complete the transformations and continue plotting, or ❌CANCEL to interrupt the generation process.
You can create a chain of these nodes by connecting the Image output to the Background input of the next node.

🛠️ Parameters

| Input/Widget | Type | Range | Default | Description | |--------------|------|-------|---------|-------------| | overlay_image | IMAGE | - | - | Foreground layer to composite | | background_image | IMAGE | - | - | Base canvas layer | | overlay_mask | MASK | - | None | Optional alpha mask (inverted on apply) | | opacity | FLOAT | 0.0 – 1.0 | 1.0 | Global transparency multiplier | | feather_type | COMBO | None / Radial In / Radial Out / Ellipse In / Ellipse Out | None | Edge softening algorithm | | blur_radius | INT | 0 – 100 | 50 | Feather intensity/blur radius | | blur_hardness | INT | 0 – 100 | 0 | Hardness of the blur borders |

💡 Tips

For quick starts use Normal Mode — all parameters are saved in the workflow
For precise tuning switch to Advanced Mode — it doesn't save state in workflow but gives more control
Corner handle scaling preserves aspect ratio
Edge handle scaling allows changing aspect ratio
If overlay has black corners after rotation, the node automatically removes them

🦊 RS RS Bypass Manager

A powerful node for managing the states of Bypass nodes and groups in complex ComfyUI circuits. If your workflow has turned into a "spaghetti monster" and you need to quickly disable entire modules (for example, switch between txt2img, inpaint and upscale), this node will save you dozens of clicks and nerves.

https://youtu.be/Ockp2SpuFSY

🔥 Features

Smart Search - Instant search for the desired nodes and groups by name right inside the drop-down menu.
Group Support - Works with ComfyUI groups. Groups can be collapsed and expanded to select individual nodes within them.
Color indication:

🔴 Red — the group or node is completely blocked.
🟠 Orange — only part of the node is bypassed in the group (partial bypass).
⚪ Gray — the node/group is active.

List of active bypasses - All blocked elements are displayed directly on the node. You can delete a bypass by clicking on the node or group name.
Smart State saving - The bypass status is saved directly in the JSON workflow. No data is lost when restarting ComfyUI, switching tabs, or sharing PNG/JSON.
Dynamic size - The node automatically adjusts its height to the number of mounted elements.
Advanced UX:

The menu does not close when you click on an item, you can quickly reset several nodes in a row.
The menu closes automatically when the mouse cursor moves outside of it (with a slight delay for comfort).
The SELECT' field... is highlighted in orange while the menu is open.
The node excludes itself from the list of elements available for bypass.

🪛 Usage

Add the 🦊 RS Bypass node to the canvas (category `🦊 RaykoStudio').
Click on the SELECT... field. A menu opens with all the groups and nodes of your scheme.
If there are a large number of nodes, use the search bar to filter.
Click on groups or nodes to switch their state (Bypass/Active).

Tip: Clicking on the name of the group bypasses it entirely. Clicking on the arrow (▶) will expand the group to select individual nodes.

When you're done, just move the mouse cursor outside the menu — it will close itself in half a second.
To remove the bypass, click on the name of the desired item in the list on the node itself.

If a node is added to a circuit that already has bypass nodes, it will automatically display them in the interface.
Also, when using bypass using comfi's own methods (the context menu is bypass, bypass in the NodeMap side menu, or bypass buttons above the node), all changes will instantly appear in the node.

🦊 RS rgb2rgba

A lightweight ComfyUI custom node that loads images while preserving the alpha channel (RGBA). Ideal for workflows that require transparency handling in PNG, WebP, and TIFF formats.

🔥 Features

Alpha Channel Preservation - Loads and outputs images with full RGBA support.
Automatic Conversion - Forces RGBA mode if the source image uses a different color mode.
Format Support - Works seamlessly with PNG, WebP, and TIFF files containing transparency.
ComfyUI Compatible - Returns a properly formatted [1, H, W, 4] float32 tensor normalized to 0.0–1.0.
Change Detection - Automatically refreshes the node output when the source file is modified.

🪛 Usage

Add the 🦊 RS rgb2rgba node to your workflow.
Upload an image or select one from your ComfyUI/input directory.
Connect the rgba output to any node that accepts ComfyUI image tensors (e.g., VAE Encode, PreviewImage, mask processors, etc.).

📝 Notes

Images without an existing alpha channel will be automatically converted to RGBA (alpha will be set to 1.0 / fully opaque).
The node expects images to be located in ComfyUI's standard input folder.
Designed to match ComfyUI's native IMAGE tensor format, ensuring drop-in compatibility with most custom nodes.

🦊 RS Any Switch

A dynamic switch node for ComfyUI that allows you to switch between multiple inputs of ANY data type with an intuitive toggle interface.

🔥 Features

Universal Type Support - Accepts any data type (IMAGE, LATENT, MODEL, AUDIO, VIDEO, TEXT, etc.)
Dynamic Inputs - Inputs are created automatically as you connect nodes (up to 20) Smart Slot Management - Empty slots are automatically removed, keeping your workflow clean
Visual Toggle Interface - Each connected input has an ON/OFF toggle switch with the source node name (multilingual)
Exclusive Selection - Only one input can be active at a time
Persistent State - Active slot selection is saved with your workflow
Auto-Numbering - Inputs are automatically numbered (Input 1, Input 2, etc.)
Smart Display - Shows connected node names with automatic truncation (max 20 chars, multilingual)
Visual Status Bar - Green/red bordered indicator showing current active slot or OFF state
'UPDATE NAME' button - Updates the names of the connected nodes, if they have changed.

🪛 Usage

Add the Node: Right-click in the graph → 🦊 RaykoStudio → RS Any Switch
Connect Inputs:
- Connect any node output to input_1
- The first connection automatically becomes active (ON)
- Additional connections create new toggle switches (OFF by default)
Switch Between Inputs:
- Click the toggle switch next to any connected input
- Turn ON the desired input (others automatically turn OFF)
- The active slot is shown in the top status bar (green border)
Disconnect Inputs:
- Simply disconnect the wire from any input
- The toggle switch is automatically removed
- Empty slots are cleaned up automatically

📝 Notes

Example 1: Switch Between Images

Load Image 1 ──→ input_1
Load Image 2 ──→ input_2
Load Image 3 ──→ input_3
                    ↓
              RS Any Switch
                    ↓
              Save Image

Toggle between different images without reconnecting wires.

Example 2: Model Switching

Checkpoint Loader A ──→ input_1
Checkpoint Loader B ──→ input_2
                           ↓
                     RS Any Switch
                           ↓
                     KSampler → VAEDecode → Save Image

Quickly compare results from different models.

Example 3: Workflow Variations

Upscale Method 1 ──→ input_1
Upscale Method 2 ──→ input_2
Upscale Method 3 ──→ input_3
                        ↓
                  RS Any Switch
                        ↓
                  Final Output

🦊 RS Outpaint

Interactive outpainting mask node with visual crop controls, preset management, and batch workflow support

🔥 Features

RS Outpaint is a custom ComfyUI node that turns mask/crop definition into an interactive, visual process. Instead of manually calculating coordinates or relying on static crops, you can drag, resize, and pan the crop region directly on the preview canvas.
The node pauses the queue until you confirm the settings, then outputs a ready-to-use control image, alpha mask, and precise dimensions. It also includes a batch preset system that remembers your crop settings and applies them automatically to subsequent images in a sequence.

❗ All changes in the node (setting the mask, creating and applying presets, replacing colors, etc. settings) are available only when it is on pause in generation - "foolproof"

Visual Mask Editor — Drag, resize, and position the crop area directly on the image preview
Pause & Approve — Workflow pauses after first execution, waiting for your confirmation
Batch Mode – Apply saved presets automatically to the next images in a queue. Auto-resets when the queue finishes
Preset System – Save, load, rename, and delete presets crop configurations.
Grid Snapping — All dimensions snap to 16px grid for clean, model-friendly outputs
Aspect Ratio Lock — Switching 🔒🔓 between maintaining crop proportions and freely choosing the mask size
Quick Presets — One-click aspect ratios: 16:9, 9:16, 21:9, 4:3, 1:1, and more
Smart Snapping — Align crop to center, edges, or fit source dimensions
Real-time UI – Zoom (scroll), pan (middle-drag), instant dimension readouts, and padding indicators
Color Mask and Background — Selecting the color for the mask and background when using the mask_image output
Padding Indicators — Visual labels show generated padding areas (▲▼◀▶)
Keyboard Shortcuts — Arrow keys to nudge crop box (Shift for 4× step)
Heartbeat — mechanism to prevent freezes
Instant reset — reset of the state after each generation

🪛 Usage

❗ The ⚙️ BATCH, ✔️ ACCEPT, and ❌ CANCEL buttons appear only after the generation is started.

Single mode
Add the node to your workflow and connect an IMAGE input.
Start the generation with the RUN button. After reaching the node, the generation will pause.
Adjust the crop area using the markers.
You can create or apply a previously created preset of settings.
Click ✔️ ACCEPT to send the configuration to the backend.
Use the outputs in your downstream nodes (ControlNet, mask blending, etc.).
If you are not satisfied with something, you can click ❌ CANCEL to interrupt the generation process.

Batch mode
Add the node to your workflow and connect an IMAGE input.
Select the number of passes in the generation queue.
Start the generation with the RUN button. After reaching the node, the generation will pause.
Adjust the crop area using the markers.
You can create or apply a previously created preset of settings.
Click the ⚙️ BATCH, it should turn green.
Click ✔️ ACCEPT to send the configuration to the backend.
Use the outputs in your downstream nodes (ControlNet, mask blending, etc.).
If you are not satisfied with something, you can click ❌ CANCEL to interrupt the generation process.

📁 Preset Storage
Presets are stored as JSON files in the presets folder within the node's directory:
ComfyUI/custom_nodes/ComfyUI_RaykoStudio/presets/
The folder is created automatically the first time you save the preset.

<details> <summary>✅ Update v0.16.3</summary>

The logic of the RS Outpaint node interface has been changed.

The default borders of the mask frame are strictly along the borders of the incoming image
All node buttons are inactive until the process queue reaches the node and is paused. The buttons become inactive immediately after the generation continues or is canceled.
All snapping presets (aspect ratio) now work without crop and without resizing the output resolution. The frame is closely adjacent to two opposite sides or to one of the sides of the image (why one? since the mask works on a 16px grid, small margins of the mask are possible, which are visually visible. This is done so that the output always results in an image with side sizes that are multiples of 16. These errors will still be masked.)
By default, the Lock/Unlock aspect ratio button is unlocked (🔓) and you can immediately start working with the mask frame without saving the aspect ratio.
When you press any snap preset button, the Lock/Unlock button is automatically locked (🔒) and all further actions with the mask frame will be performed while maintaining the selected aspect ratio until you unlock the button again.

</details>

💡Tip: For precise positioning, use the arrow keys on your keyboard (hold Shift to move faster)

↔️ Inputs and Outputs:

Input :
image - Source image for outpainting

Output :
control_image - Image with masked area (gray padding where generation will occur)
control_mask - Binary mask: 0 = keep original (black), 1 = generate new content (white)
mask_image - Makes an image from a mask
width - Final output width (after optional resize)
height - Final output height (after optional resize)

🦊 RS Ref 2 Latent

A lightweight node that encodes a reference image into latent space and injects it into positive/negative conditioning for reference-based generation workflows.

🔥 Features

Image Processing - Takes reference image and encodes it through VAE
Conditioning Injection - Automatically adds reference latents to both positive and negative conditioning
Automatic Size Handling - Intelligently scales image dimensions to match VAE requirements
Clean & Simple - No unnecessary complexity - just reference image → latent → conditioning
Compatible - Works with all VAE-based models

↔️ Inputs and Outputs

| Input | Type | Description | |-------|------|-------------| | vae | VAE | VAE model used to encode the image into latent space | | image | IMAGE | Reference image to encode | | positive | CONDITIONING | Positive conditioning from CLIP text encoder | | negative | CONDITIONING | Negative conditioning from CLIP text encoder |

| Output | Type | Description | |--------|------|-------------| | positive | CONDITIONING | Modified positive conditioning with reference latents | | negative | CONDITIONING | Modified negative conditioning with reference latents | | latent | LATENT | Encoded reference image as latent |

🪛 Usage

Input Processing: The node receives your reference image and VAE
Size Calculation: Automatically calculates optimal dimensions (multiples of 64 for most VAEs)
Encoding: Upscales the image using Lanczos interpolation and encodes it through VAE
Conditioning Injection: Adds the encoded latent to both positive and negative conditioning under the key reference_latents
Output: Returns modified conditioning and the latent representation

🦊 RS Image Compare

A node that provides an interactive image comparison interface with zoom and pan controls

🔥 Features

Side-by-side comparison - Compare two images using an interactive slider
Zoom control - Zoom in from 1.0x to 10.0x with precise control
Pan controls - Pan horizontally and vertically to explore zoomed images (-100% to +100%)
Interactive slider - Drag the divider to reveal more of either image
Reset functionality - Individual reset buttons for each parameter + global "Reset All" button
Auto-reset - Parameters automatically reset to defaults when new images are loaded

🪛 Usage

Slider - Drag the vertical divider left or right to compare images

Left position: Shows only image_2
Right position: Shows only image_1
Middle position: Shows both images equally

Zoom - Adjust zoom level from 1.0 to 10.0

Drag the slider or click anywhere on the slider track to jump to that position
Click the 🔃 button to reset to 1.0

Pan H - Horizontal panning from -100% to +100%

Works in conjunction with zoom to navigate the image
Click the 🔃 button to reset to 0

Pan V - Vertical panning from -100% to +100%

Works in conjunction with zoom to navigate the image
Click the 🔃 button to reset to 0

🔄 Reset All Parameters - Reset all controls (Zoom, Pan H, Pan V) to their default values

⚙️ Technical Details

All pan and zoom operations are performed relative to the image center
Pan values are expressed as percentages of the zoomed image dimensions
Images are automatically scaled to fit the display area while maintaining aspect ratio

🔎 Use Cases

Before/after image comparisons
Quality assessment between different models or configurations
Detailed inspection of image differences at various zoom levels

🦊 RS Load Image

A node for load image and creating a spline mask

🔥 Features

The node has the functionality of the native Load Image node, but instead of the native mask editor, a built-in spline editor is used. More accurate selection based on the principle of the Lasso tool from Photoshop.

🪛 Usage

The node is ready for use immediately after it is added. Images are added using the "🎨 IMAGE" (images from the input folder) and "🖼️ UPLOAD IMAGE" (images from any folder on your PC) buttons. You can scale the node to a convenient size to more accurately place the points of the spline. Incorrectly positioned points can be deleted by left-clicking with the CTRL key held down. To remove all points from the preview area, click the "🔴 CLEAR POINTS" button.

↔️ Inputs and Outputs:

The IMAGE output returns the original image unchanged. The MASK output returns a black and white mask where the white area corresponds to the drawn polygon.

🦊 RS Intermediate Spline Mask

An interactive node for creating intermediate spline masks

🔥 Features

Node allows you to pause the workflow at the image processing stage, manually select the desired area and continue generation without completely restarting the process. More accurate selection based on the principle of the Lasso tool from Photoshop.
The ⚙️ BATCH button enables continuous processing mode, allowing the node to reuse the same mask across multiple queue iterations without requiring manual re-drawing.

🪛 Usage

When this node is reached, pipeline execution is automatically suspended. A preview of the input image is displayed in the node's interface. The user can left-click to add polygon points. Incorrectly positioned points can be deleted by left-clicking with the CTRL key held down.
The "✔️ ACCEPT" button confirms the created mask, after which the node completes processing and transmits the data further according to the scheme. The "🔴 CLEAR POINTS" button clears all the drawn points for redrawing. It is important that after using this button, you do not need to press the Prompt Queue again, just draw a new mask and press "✔️ ACCEPT". The "❌ CANCEL" button completely interrupts the process and resets the node status.

Single Run (Default)

Queue prompt → node pauses, shows image
Draw mask on overlay
Click ✔️ ACCEPT
Processing continues normally

Batch Run

Set the value for batch processing in the Batch Count menu (located next to the RUN button)
Queue prompt → node pauses, shows image
Draw mask on overlay
Click ️ BATCH (button turns orange)
Click ✔️ ACCEPT — mask is saved, queue continues
On next iteration: no pause, saved mask is applied automatically
Repeat until Batch Count is exhausted or queue is stopped

Stopping Batch Mode
Batch mode is automatically deactivated in the following cases:

❌ CANCEL button is clicked — stops the queue and clears the saved mask
Queue completes — after the last iteration finishes, batch mode resets automatically (~2 seconds delay)
Node is removed from the graph — saved mask is cleared

↔️ Inputs and Outputs:

The IMAGE input accepts an image from any previous node. The IMAGE output returns the original image unchanged. The MASK output returns a black and white mask where the white area corresponds to the drawn polygon.

🦊 RS Image Selector

Node for Interactive Batch Image Selection

Screenshot_2

🔥 Features

Interactive Grid View - Display all batch images in a responsive grid layout
Multi-Select Support - Click to select/deselect individual images
Smart Auto-Resize - Node automatically adjusts size based on image count
Heartbeat System - Robust connection monitoring between frontend and backend
Auto-Cleanup - Proper resource cleanup on node removal or workflow close

🪛 Usage

Buttons
➕ SELECT ALL - Select all images in batch
⭕ DESELECT ALL - Clear all selections
✔️ ACCEPT - Confirm selection and continue
❌ CANCEL - Cancel and interrupt generation

Continue Workflow
Click "✔️ ACCEPT" to pass selected images to next nodes. Only selected images will be processed downstream

⚠️ Reminder

The generation process will pause indefinitely until you click "✔️ ACCEPT", "❌ CANCEL" or close the workflow or the ComfyUI page.

🦊 RS MultiLatent

Universal latent generation node - automatically adapts to any VAE architecture.

🔥 Features

Automatic VAE Detection - reads latent_channels and scale_factor directly from VAE
Universal Compatibility - works with SD1.5, SDXL, Flux, Flux.2, Krea2, QwenImage, and many other models
Flexible Sizing - three modes: Preset, Custom, and Megapixels
100+ Presets - extensive library of ready-to-use resolutions
Smart Device Handling - automatically matches VAE device and dtype

🪛 Usage

Connect VAE: Plug any VAE into the vae input
Choose size mode:

Preset: Select from 100+ predefined resolutions
Custom: Manually enter width and height
Megapixels: Specify target resolution in MP with aspect ratio

Connect to sampler: Use the latent output with KSampler or other nodes

⚙️ Settings

size_mode: Size definition mode

Preset - Use predefined resolutions
Custom - Manual width/height input
Megapixels - Target resolution in megapixels

preset (only for Preset mode): Choose from 100+ resolutions including:

Square formats (512×512 to 2016×2016)
Portrait formats (4:5, 3:4, 2:3, 9:16, 21:9, etc.)
Landscape formats (5:4, 4:3, 3:2, 16:9, 20:9, etc.)

width/height (only for Custom mode): Custom dimensions (8-4096px)

megapixels (only for Megapixels mode): Target resolution (0.1-10.0 MP)

aspect_ratio (only for Megapixels mode): 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16, 21:9, 9:21, 4:5, 5:4

batch_size: Number of latent images to generate (1-64)

🦊 RS Image to Latent & 🦊 RS Image to Latent (simplified)

A powerful and user-friendly ComfyUI node that converts images to latents with intelligent size optimization.

🦊 RS Image to Latent

🔥 Features

Multiple sizing modes - Auto, Preset, Custom, or Megapixels
VAE-aware - Automatically detects divisibility requirements
Batch processing - Create multiple identical latents at once
Smart upscaling - Choose from multiple upscale methods
Clean interface - No technical clutter, just what you need

🪛 Usage

Resolution control - Ensure consistent sizes across your workflow
VRAM optimization - Use megapixels mode to stay within limits
Auto mode - Preserves original proportions
Batch generation - Create multiple variations from one image

🎯 Modes Explained

Auto
Preserves original image size and only rounds to meet VAE divisibility requirements. No unexpected upscaling.

Preset
Choose from 50+ common resolutions including:

Square: 512×512 to 1920×1920
Portrait: 3:2, 3:4, 9:16 and more
Landscape: 4:3, 16:9, 21:9 and more

Custom
Manually enter width and height (must be multiples of 8).

Megapixels
Set target megapixels with specific aspect ratio. Perfect for controlling VRAM usage.

🔧 Rounding Modes

auto - Picks the nearest valid size (recommended)
shrink - Only reduces size, never increases
expand - Only increases size, never reduces

↔️ Input and output:

Input:
image - Input image to convert to latent
vae - VAE model for encoding (auto-detects divisibility)

Output:
latent - The encoded latent tensor
width_px - Final width in pixels
height_px - Final height in pixels
width_latent - Latent width (width/8)
height_latent - Latent height (height/8)

🦊 RS Image to Latent (simplified)

A simplified node that converts images to latent with intelligent size optimization.

🔥 Features

Minimum settings - Only batch size
Default settings:
Automatically determines the multiplicity from VAE
Rounds the image size to a multiple of the value
If the size has changed, it will be resized via lanczos
If it is already multiple, it encodes directly without a recycle
Supports batch_size to create a batch of identical latents
Returns 3 outputs: latent, width_px, height_px

🦊 RS Crop Image and 🦊 RS Insert Crop

An interactive node that allows you to visually crop an image directly within the node interface. Unlike standard Crop nodes, here you see the image and draw the crop rectangle with your mouse — making the process precise and intuitive.
The main feature is the Multiple mode, which guarantees that the cropped image size will be a multiple of a specified number (4, 8, 16, 32, or 64). This is critical when working with VAEs and generative models (SDXL, FLUX, SD 1.5б etc.) that expect certain size multiples.
With further insertion of the cut fragment after its modification back into the original image.
Automatically detect and crop regions of interest using masks. When a mask is connected, the node analyzes it and sets the crop area to the bounding box of all non-zero pixels.

Usage only image

Usage with mask

Usage Insert Crop

Screenshot_2

🔥 Features

Interactive cropping — the image is displayed directly in the node, the rectangle is drawn with the mouse
Moving - Dragging the selected area
Process pause - Workflow waits for confirmation before continuing
Visualization - Real-time area size display
Three actions — Accept, Reset, Cancel
Smart alignment — Automatic size snapping to a chosen multiple
Reverse paste support — The CROP_DATA output contains precise coordinates for a second node that can paste the crop back into the original image
State persistence — Rectangle parameters are saved in the workflow
Boundary protection — The rectangle never leaves the image bounds
Automatic Bounding Box Detection — Calculates crop coordinates from the extreme points of the mask (leftmost, rightmost, top, bottom)
Smart Thresholding — Uses a 0.5 threshold to determine active mask pixels
Seamless Integration — Works with existing multiple_mode for grid-aligned cropping
Visual Feedback — Crop rectangle appears automatically on the overlay when mask is connected

🦊 RS Insert Crop is a node that allows you to seamlessly insert a previously cut fragment back into the original image after it has been modified by the workflow nodes.

Crop Data - Inserting a fragment into the original image using the previously obtained precise cutout parameters.

🪛 Usage only image

🦊 RS Crop Image
Connect the image to the IMAGE input.
Run queue (queue request) — the node will pause operation and display the clipping area as a rectangle.
Adjust the rectangle borders:

Drag the rectangle — move it entirely
Drag the corner markers — resize the rectangle

If necessary turn on "Multiple" and select a multiplier.
Click ✔️ ACCEPT button — the node will continue the queue operation and output the cropped image.

🪛 Usage with mask

Connect your image to the image input
Connect a mask to the mask input
The crop area will automatically adjust to fit the mask boundaries
If multiple_mode is ON, the crop box will be aligned to the specified multiple (8, 16, 32, 64, etc.)
Click ✔️ ACCEPT to apply the crop or adjust manually

Button:
✔️ ACCEPT - Confirm the allocation and continue workflow
🔄 RESET - Reset the selection and select again
❌ CANCEL - Cancel and interrupt generation

🦊 RS Insert Crop
Connect the original image to the Original Image input.
Connect the embedded and modified fragment to the Cropped Image input.
Connect the node with the text Crop Data.

🎯 Use Cases

1. Preparing a crop for VAE/generation
Enable Multiple = ON, choose 8 (for SD 1.5 / SDXL) or 16/64 (for FLUX). The result is guaranteed to be accepted by the VAE without warnings about non-multiple sizes.
2. Precise region editing
Crop the desired area → send it to an inpaint node → after generation, paste the result back using CROP_DATA. The reverse paste will be pixel-perfect.
3. Comparing results on the same region
Draw a rectangle, save the parameters → run different models/seeds through the same crop → CROP_DATA stays consistent between runs.
4. Isolating a face/object
Scale up the node, draw a precise rectangle around a face, send it to a face-swap or restoration pipeline.

⚠️ Reminder

The generation process will be suspended indefinitely until you click "✔️ ACCEPT", "❌ CANCEL" or close the workflow or the ComfyUI page.

🦊 RS Styles Loader

Node for uploading and managing styles from CSV files

Screenshot_1

🔥 Features

Download CSV files - Upload your files with styles directly through the interface
Tree structure - Styles are organized by folders (categories)
Visual selection - User-friendly interface with drop-down lists
Bypass styles - Turn styles on/off without deleting them from the list
Combining - Multiple styles are combined into one prompt
Save to workflow - All settings are saved along with the project

🪛 Usage

Buttons
SELECT CSV FILE - Select the desired CSV file from the list of files that you have already uploaded earlier.
📂 UPLOAD NEW CSV FILE - Download a new CSV file from anywhere on your PC. It will automatically appear in the styles folder and in the future you will be able to select it with the "SELECT CSV FILE" button.
➕ ADD STYLE - Choosing the styles you need
🟢 - Bypass on/off. You can choose an infinite number of styles and change them to create the combination you need.
❌ - Removing a style from the panel

🎨 Styles

You can use your own styles or find them in the ComfyUI community
For example:
https://github.com/vaulthunt3r/ComfyUI-Style-Prompts-Collection
https://github.com/Art-xmaster/comfyui-AGSoft/tree/main/styles

⚠️ Notes

The styles folder is created automatically at the first startup.
When deleting a node from workflow, the uploaded CSV files are not deleted.
To update the list of files after adding CSV manually, restart ComfyUI.

🦊 RS Last Frame

A lightweight ComfyUI node that extracts the last frame from any video input.

Screenshot_5

🔥 Features

Universal Input Support — Accepts native VideoFromFile objects (ComfyUI's built-in video loader), standard IMAGE tensors (VHS, Image Batch), and dictionary formats
Zero Dependencies — Works out-of-the-box without requiring Video Helper Suite or other external libraries
Automatic Format Detection — Intelligently handles different tensor dimensions [F, H, W, C], [B, F, H, W, C], or channel-first formats
Memory Efficient — Extracts frames without unnecessary copying or conversion

🪛 Usage

If video has only 1 frame, that frame is returned unchanged
Automatically handles batched inputs by taking the first batch element
Channel permutation applied automatically if decoder outputs [F, C, H, W]
Compatible with ComfyUI v1.0+ and all major video loading extensions

↔️ Input and output:

Input: video_frames (* — Universal)
Accepts:

IMAGE tensor from VHS loaders, Image Batch, or standard image nodes
VideoFromFile object from ComfyUI's native Load Video node
Dictionary with frames, images, or video keys

Output: IMAGE (torch.Tensor)
Single frame tensor with shape [1, H, W, C] — ready for VAE encoding, preview, or any image processing node.

🦊 RS Models Loader

A powerful universal model loading node that combines UNET, CLIP (with additional dual-clip support), VAE and LoRa downloads in a single interface and the function of saving model configurations to preset presets.

🔥 Features

Unified Loading - Load UNET, CLIP, VAE, and multiple LoRAs in one node
Dual CLIP Support - Toggle between single CLIP mode (CLIPLoader) and dual CLIP mode (DualCLIPLoader) for models like Flux, SD3, Hunyuan DiT, and PixArt
LoRA Management - Add, enable/disable, and adjust strengths for multiple LoRAs with an intuitive visual interface
Persistent Storage - LoRA configurations are saved per node and persist across sessions
Folder Structure - Browse LoRAs with folder tree navigation
Search Functionality - Quickly find LoRAs by name
Presets of model sets - Save model sets to presets for quick switching between models

📃 Notes

When dual CLIP mode is disabled, clip_name2 is visible but disabled (grayed out)
The node automatically falls back to single CLIP mode if clip_name2 is empty
LoRA strengths can be set independently for model and CLIP
All settings are saved with your workflow and persist across ComfyUI sessions

🦊 RS Prompts

Node that provides enhanced prompt management with visual controls, pause-for-edit mode, and external input toggling

🔥 Features

Dual prompt sources - Use internal textarea or external text input
External input toggle - Enable/disable external text input with one click
Pause-for-edit mode - Interrupt generation, edit prompts, and continue
Visual status indicator - Shows current prompt source and mode
Prompt presets - Save, load, and delete prompt templates
Clean interface - Minimal design that blends with ComfyUI

↔️ Input and output:

Inputs

| Name | Type | Description | |------|------|-------------| | clip | CLIP | CLIP model for encoding | | text_input | STRING (optional) | External prompt source |

Outputs

| Name | Type | Description | |------|------|-------------| | POSITIVE | CONDITIONING | Encoded positive prompt | | NEGATIVE | CONDITIONING | Encoded negative prompt (empty) | | PROMPT_STRING | STRING | Final prompt text |

🪛 Usage

Prompt Sources

Local prompt: Uses text from the internal textarea
External input: Uses text from connected text_input node

The 🔘 Disable text input toggle lets you temporarily ignore external input without disconnecting cables.

Pause Mode

Enable ⏸️ Pause for edit
When external data arrives, the node pauses and shows an overlay
Edit the prompt as needed
Click APPROVE (use edited) or REJECT (use original)
Generation continues with your choice

The pause toggle stays enabled after approval, automatically pausing on next external data.

Usage Examples

Local Prompt Only

Leave text_input disconnected
Or connect but enable 🔘 Disable text input
Write prompts directly in textarea

External Prompt Source

Connect a text node to text_input
Disable 🔘 Disable text input
Enable ⏸️ Pause for edit for manual review

Prompt Presets

Write your prompt
Click 💾 Save prompt and enter a name
Later click 📂 Select prompt to load
Delete presets via ❌ button in the list

🖥️ Node Interface

| Control | Description | |---------|-------------| | 🔘 Disable text input | Toggle external input on/off (🔴 when disabled) | | ⏸️ Pause for edit | Enable edit mode (activates when external data arrives) | | Status indicator | Shows: 📝 Local prompt / 🔌 External input / ⏸️ WAITING FOR EDIT | | Text area | Main prompt editor | | ❌ Clear prompt | Clear the text area | | 💾 Save prompt | Save current prompt as preset | | 📂 Select prompt | Load saved prompt preset | | ✔️ APPROVE & CONTINUE | Accept edited prompt and continue | | ❌ REJECT | Keep original prompt and continue |

📂 Prompts Storage

Prompts are stored as JSON files in the prompts folder within the node's directory:
ComfyUI/custom_nodes/ComfyUI_RaykoStudio/prompts/
The folder is created automatically the first time you save the prompt.

🐛 Known Issues & Solutions

Text not saving - Check write permissions for prompts folder
Pause mode not working - Ensure toggle is ON before starting generation

🦊 RS Text Overlay Pro

Interactive ComfyUI node for overlaying text on images with real-time positioning, scaling, rotation, and rich text effects

🔥 Features

Interactive canvas editor — drag, scale, and rotate text overlay directly on the image
Rich text effects — Outline, Glow, and Shadow with full parameter control
Hybrid rendering — instant client-side preview + high-quality server-side final render
Font system — custom font library with search and live preview
Smart UI — collapsible sidebar sections with per-effect toggles
Dual modes — compact widget mode inside the node + full-screen Advanced Mode editor
Color tools — HEX input with validation + native color picker
Multiline support — textarea with line spacing and letter spacing controls
Text alignment — left / center / right with visual icons
State isolation — each effect can be enabled/disabled independently; disabling resets parameters to defaults

By default, the node is displayed as a standard node with a thumbnail and a minimal set of widgets.
All changes are immediately displayed.
Clicking on the "Advanced Mode" button opens a full-screen editor where you can visually manipulate the text.

Advanced Mode:

The canvas occupies the main part of the screen, with the settings panel on the left/right.
The text can be dragged, resized by corners and sides, and rotated using a special pen.
All changes are immediately displayed.

The control panel has sections: TEXT, OUTLINE SETTINGS, GLOW SETTINGS, SHADOW SETTINGS. Each section is collapsible. Controls: text field, sliders, color pickers, alignment and font selection buttons.

Interactive elements: they are highlighted when hovering, and there are pop-up windows for precise input for numeric values.

###❗Requirements
It requires the installation of the pycairo library (just install the dependencies requirements.txt )
The library is needed for high-quality text rendering and effects.

🔤 Installing Fonts

Place .ttf, .otf, or .ttc files in folder:
ComfyUI\custom_nodes\ComfyUI_RaykoStudio\fonts
Restart ComfyUI
Fonts are auto-detected on editor open. Arial.ttf is used as fallback.

🚀 Usage

Add 🦊 RS Text Overlay Pro node to your workflow
Connect an image to the image input
Run the queue
Configure text, effects, and position
Click ✔️ APPLY to render the final image

Quick start

Type your text in the TEXT section
Pick a font and adjust size via the canvas handles
Enable effects via toggles in OUTLINE / GLOW / SHADOW sections
Drag the green bounding box to position the text, turn the image by the orange rotation knob
Click ✔️ APPLY when done

⚙️ Parameters

TEXT

| Parameter | Type | Default | Description | |---|---|---|---| | Text | textarea | "" | Multiline text content | | Font | select | first in library | Font family | | Text Color | color | #FFFFFF | Fill color | | Text Opacity | slider | 1.0 | 0.0 – 1.0 | | Line Spacing | slider | 1.0 | 0.5 – 3.0 | | Letter Spacing | slider | 0.0 | -20 – 100 | | Alignment | buttons | center | Left / Center / Right |

OUTLINE SETTINGS

Enabled automatically when Thickness > 0.

| Parameter | Type | Default | Description | |---|---|---|---| | Thickness | slider | 0 | 0 – 50 px | | Color | color | #808080 | Outline color | | Opacity | slider | 1.0 | 0.0 – 1.0 |

GLOW SETTINGS

Toggle ENABLE GLOW to activate.

| Parameter | Type | Default | Description | |---|---|---|---| | Glow Color | color | #FFFFFF | Glow color | | Size | slider | 100 | 0 – 200 px | | Spread | slider | 150 | 0 – 300 px | | Opacity | slider | 1.0 | 0.0 – 1.0 |

SHADOW SETTINGS

Toggle ENABLE SHADOW to activate.

| Parameter | Type | Default | Description | |---|---|---|---| | Shadow Color | color | #333333 | Shadow color | | Offset X | slider | 10 | -30 – 30 px | | Offset Y | slider | 10 | -30 – 30 px | | Blur | slider | 15 | 0 – 100 px | | Opacity | slider | 0.8 | 0.0 – 1.0 |

️ Controls

Normal Mode (inside node)

🔍 ADVANCED MODE button — opens full-screen editor
✔️ APPLY — renders and closes editor
❌ CANCEL — discards changes

Advanced Mode (full-screen editor)

| Action | Control | |---|---| | Move text | Drag inside bounding box | | Scale | Drag corner/edge handles | | Rotate | Drag orange handle above box | | Zoom canvas | Mouse wheel | | Close editor | Esc key | | Toggle section | Click section header | | Enable effect | Click toggle in header |

🦊 RS Color Picker

Professional color picker node with advanced features including eyedropper and color history

🔥 Features

Visual Color Picker - Intuitive color selection with live preview
Eyedropper Tool - Pick colors from anywhere on screen (Chrome/Edge) or from ComfyUI canvas (all browsers)
Copy to Clipboard - Double-click color swatch to copy HEX value
Basic Colors - 8 preset colors for quick access
Recent Colors - History of last 24 used colors (saved in localStorage)
Multiple Outputs - HEX_INT, HEX_STR, and RGB formats
Clean UI - Professional interface matching ComfyUI design

🪛 Usage

Select a color using one of these methods:

Click the color swatch to open native color picker
Click the palette icon (🎨) to activate eyedropper tool
Click any preset color from the "Basic Colors" row
Click any color from the "Recent Colors" history
Type HEX value directly into the input field

Connect outputs to other nodes:
HEX_INT - Integer value (e.g., 16711680 for #FF0000)
HEX_STR - HEX string (e.g., "#FF0000")
RGB - Normalized RGB values (e.g., "1.000, 0.000, 0.000")

🦊 RS Saturation

Professional image saturation control with artifact and highlight protection.

🔥 Features

Smooth adjustment with 0.05 steps
Smart boosting without overexposure
Artifact protection even at extreme values
Batch processing optimized

🪛 Usage

RS Safe Saturation

| Range | Processing Type | Use Case | |------------|-------------------------------|-----------------------------| | 0.0-0.9 | Toning/desaturation | Gradual color removal | | 1.0-1.3 | Natural enhancement | Recommended range | | 1.3-2.0 | Vibrant artistic effects | Stylization | | 2.0-3.0 | Maximum saturation | Cinematic effects |

⚙️ Technical Details

Algorithm workflow:
Luminance space conversion
Non-linear adjustment:
Values <1.0: Linear interpolation
Values >1.0: Adaptive S-curve
Auto highlight recovery

🦊 RS Save Image

Node for adding explanatory text to an image

Screenshot_7

🔥 Features

The node is used to save the image while preserving the workflow inside the image. You can add explanatory text to an image with a choice of background size, theme, font and its size. It is possible to add your own fonts (to the fonts folder). Use ttf and otf fonts.
If the label is not needed, leave the text field blank and the image is saved as usual.
Themes:
light - white background, black text.
dark - black background, white text.

🦊 RS Save Image Pair

The node is used to save the original and final images in a single image, while maintaining the workflow within the image

Screenshot_1

🔥 Features

The node is used to save the source and final images in a single image while maintaining the workflow within the image. You can add explanatory text to any image with a choice of background size, theme, font and size font. A reverse upscale from 1 to 0 is provided to reduce the saved image (if it is used as a sketch with a workflow inside).
It is possible to add your own fonts (to the fonts folder). Use ttf and otf fonts.
If the label is not needed, leave the text field blank and the image is saved as usual.
The node is convenient for visual understanding of the workflow contained in the image.

🪛 Usage

It is better to choose horizontal saving for portraits, and vertical saving for landscapes.
Themes:
light - white background, black text.
dark - black background, white text.

🦊 RS Image-Text

Node embeds any hidden text into the image that can be used later

RS Image-Text

🔥 Features

The node writes any hidden text to any png and jpeg file (jpeg is converted to png). And outputs text from images recorded in this way. You can use it instead of a Load Image (without a mask) and transfer the recorded text to the promt node. It is useful if there is an image and a prompt to it, but the image does not contain a workflow (often found on the site civitai.com).

🪛 Usage

Two modes:
Write - writes text to the uploaded image and saves it to the output folder with the prefix you specified.
Read - reads the text you wrote earlier in the uploaded image, sends the text and images further according to the scheme.

Link to the video: https://youtu.be/1s26hUcVXX4

🦊 RS Loop Switch

A combined node for generating a sequence of values with automatic switching

Screenshot_1

🔥 Features

The node generates a list of INT values, automatically switching between 10 preset values based on the current cycle step. It is ideal for generating a series of images with different seeds without the need to manually change the parameters. But you can use it for any other tasks where dynamic INT changes are required.

🔌 Connection (examples)

RS Loop Switch (output) → KSampler (seed)
RS Loop Switch (output) → KSampler (steps)
RS Loop Switch (output) → any INT input

🦊 ComfyUI Settings Manager

A user-friendly extension that allows you to easily backup, restore, and manage your interface settings. Never lose your custom UI layout, preferences, or configurations again!

🔥 Features

Smart Backups - Save your current interface settings with custom names. If no name is provided, it automatically uses a timestamp.
Easy Restore - Select any previous backup from a clean UI list and restore it in one click.
Built-in Server Restart - Restoring settings requires a server restart. The extension provides a dedicated, pulsing RESTART SERVER button to make this obvious and safe.
Auto-Refresh Polling - After restarting, the extension automatically pings the server and refreshes your browser page the moment ComfyUI is back online.
Manage Backups - Delete old or unnecessary backups directly from the interface without leaving ComfyUI.
Safe Naming - Automatically sanitizes custom backup names (removes invalid characters) and prevents overwriting by adding suffixes (_1, _2) if a name collision occurs.
You can use common settings for multiple versions or assemblies of ComfyUI.
You can create several interface configurations and change them with a couple of clicks, rather than crawling through all the settings tabs.
It's also useful for those who mess up the interface settings a lot - you can always roll back to the previous version.

🪛 Usage

Accessing the Manager
Look for the Desktop/Monitor icon (🖥️) in the left sidebar of ComfyUI. Click it to open the Settings Manager panel.

Saving Settings (Backup)

In the Save Interface Settings section, you will see a text field.
(Optional) Enter a custom name for your backup (e.g., my_favorite_layout).
- Note: If you leave it empty, the current date and time will be used.
Click the 💾 Save Settings button.
A green status message will appear at the bottom confirming the save. If the list of backups is open, the new backup will instantly appear at the top.

Restoring Settings

In the Restore Interface Settings section, click 📂 Load Backups.
A list of your saved backups will appear.
Click ✓ Restore next to the backup you want to load.
The button will change to ✓ Restored, and a new, pulsing orange button 🔄 RESTART SERVER will appear.
Click 🔄 RESTART SERVER.
The extension will automatically wait for the server to come back online and refresh your browser page automatically.

Deleting Backups

Load your backups list.
Click the red ✕ button next to any backup you want to remove.
The backup is instantly deleted from the UI and the file system.

📁 File Structure

All backups are stored outside the ComfyUI directory to keep your installation clean. By default, they are saved in your user Documents folder:

📂 Documents/
 ── 📂 ComfyUI_Settings_Backups/
      ├── 📂 2026-06-09_12-40-44/
      │    ── 📄 comfy.settings.json
      ├── 📂 my_favorite_layout/
      │    └── 📄 comfy.settings.json
      ── ...

⚙️ Technical Notes

Server Restart - The extension uses os.execv to restart the Python server. If you are using a third-party launcher (like Stability Matrix or Pinokio), the server might close but not automatically reopen. In this case, simply restart your launcher manually.
Polling - After a restart, the extension pings the server every 5 seconds (up to 10 times) to detect when it's fully loaded before triggering the page reload.

🤝 Bug Reporting

If you encounter an issue or find a bug:

Check Issues section on GitHub, maybe problem is already known.
If new problem, create new Issue describing:

ComfyUI and Python versions
Problem description and reproduction steps
Screenshots or error logs (if available)

📜 License

Apache License 2.0. Use at your own risk without any warranties. See the LICENSE file for details

❤️ Acknowledgments

Thanks to ComfyUI community for inspiration and support.
Special thanks to FotoSHAMAN for his fantastic ideas!
If you like this node, don't forget to star on GitHub!

Run ComfyUI workflows without the setup

No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

Learn more

ComfyUI Extension: ComfyUI_RaykoStudio

Custom Nodes (26)

README

🦊 ComfyUI_RaykoStudio

Set of custom nodes for ComfyUI providing additional image processing capabilities

📃 Requirements

🛠 Installation

🦊 RS Collage Node

🔥 Features

🖼️ Working Modes

🔆 Feathering Types

🪛 Usage

🛠️ Parameters

💡 Tips

🦊 RS RS Bypass Manager

🔥 Features

🪛 Usage

🦊 RS rgb2rgba

🔥 Features

🪛 Usage

📝 Notes

🦊 RS Any Switch

🔥 Features

🪛 Usage

📝 Notes

🦊 RS Outpaint

🔥 Features

🪛 Usage

↔️ Inputs and Outputs:

🦊 RS Ref 2 Latent

🔥 Features

↔️ Inputs and Outputs

🪛 Usage

🦊 RS Image Compare

🔥 Features

🪛 Usage

⚙️ Technical Details

🔎 Use Cases

🦊 RS Load Image

🔥 Features

🪛 Usage

↔️ Inputs and Outputs:

🦊 RS Intermediate Spline Mask

🔥 Features

🪛 Usage

↔️ Inputs and Outputs:

🦊 RS Image Selector

🔥 Features

🪛 Usage

⚠️ Reminder

🦊 RS MultiLatent

🔥 Features

🪛 Usage

⚙️ Settings

🦊 RS Image to Latent & 🦊 RS Image to Latent (simplified)

🦊 RS Image to Latent

🔥 Features

🪛 Usage

🎯 Modes Explained

🔧 Rounding Modes

↔️ Input and output:

🦊 RS Image to Latent (simplified)

🔥 Features

🦊 RS Crop Image and 🦊 RS Insert Crop

Usage only image

Usage with mask

Usage Insert Crop

🔥 Features

🪛 Usage only image

🪛 Usage with mask

🎯 Use Cases

⚠️ Reminder

🦊 RS Styles Loader

🔥 Features

🪛 Usage

🎨 Styles

⚠️ Notes

🦊 RS Last Frame

🔥 Features

🪛 Usage