A collection of nodes that provide dynamic dropdown selectors for Samplers, Schedulers, Checkpoints, and Diffusion Models, outputting a comma-separated string for use in XY plots.

DemonAlone-nodes-ComfyUI

A flexible pass-through node that accepts any input type. It safely forwards data downstream or returns a clean None output if the input is unconnected, ensuring pipeline stability without breaking custom workflows.

AnyAdapterNode

Concatenates any number of up to 5 text inputs into a single string using a custom delimiter. Acts as a flexible joiner that automatically ignores unconnected slots, ideal for building dynamic text prompts or combining parameters from various sources.

AnyConcatNode

Safely converts any compatible input value into a floating-point number. If the conversion fails or is impossible, it returns None to prevent workflow errors.

AnytoFloatAdapterNode

Safely converts any compatible input value into an integer. If the conversion fails or is impossible, it returns None without crashing the workflow.

AnytoIntegerAdapterNode

A conditional routing node that outputs either the 'true' or 'false' input value based on the current boolean state.

BooleanSwitchNode

Retrieves checkpoint model names from the ComfyUI checkpoints folder and outputs both the filename tuple (for combo menus) and its string representation, allowing dynamic model selection in utility chains.

CheckpointSelectorNode

Outputs an integer clip skip value ranging from -24 to -1. Provides fine-grained control over the depth of CLIP token skipping in diffusion models.

ClipSkipSliderNode

Decodes latents to images using the provided VAE when enabled, outputting both Latent and Image for previewing. When disabled, acts as a pass-through (Reroute) node, passing the original Latent unchanged without decoding.

ConditionalVAEDecodePreview

Calculates and returns the total number of elements in an input list. Provides a quick way to determine batch sizes, list lengths, or count available inputs within your workflow logic.

CountListNode

Outputs audio parameters: number of channels, sample rate, duration, tensor shape.

DA_AudioDebugNode

Executes standard diffusion sampling using the vanilla KSampler implementation

DA_Base_KSampler

Aggregates up to 10 connections into a single bus pipe. Preserves positions of empty inputs.

DA_BusInNode

Unpacks the bus pipe back into up to 10 outputs.

DA_BusOutNode

An advanced KSampler supporting Model Shift (SDE), dynamic noise masking, and customizable preview frequency. Allows users to control how often latent previews are rendered during the sampling process to optimize memory usage and visual feedback.

DA_Enhanced_KSampler

Loads previously saved latent tensors (.latent or .safetensors) from the current output directory. Automatically detects and parses both modern JSON-wrapped formats (supporting float32 tensors) and legacy binary formats (pickle/torch/safetensors). Ensures 4D tensor shape compatibility by broadcasting missing dimensions. Logs loaded tensor statistics (shape, min/max/mean) for debugging.

DA_LatentLoader

Tiled diffusion upscaling with optional ESRGAN pre-scaling. Memory-efficient processing of large images in tiles, featuring seamless overlap blending, configurable denoise strength, and preview callbacks for detail enhancement without VRAM exhaustion.

DA_TiledUpscaler

Collects custom or pre-processed diffusion model file paths from multiple sources (up to 10 inputs). Outputs a consolidated list and string representation for seamless integration into multi-model workflows.

DiffusionModelGeneratorNode

DiffusionModelSelectorNode

A simple utility node that outputs a single float value with adjustable range and precision. Useful for creating custom selectors like CFG strength.

FloatSelectorNode

A versatile utility node for synchronously resizing images and masks. Supports multiple scaling methods (stretch, keep proportion, fill/crop, pad), various interpolation filters, and conditional logic to resize only when necessary. Automatically handles empty outputs if an input tensor is missing.

1 = disable, otherwise round down to multiple

ImageResizeNode

Merges two input lists into a single combined list, preserving the order of elements from both sources. Essential for chaining multiple generated lists (e.g., samplers or encoders) into one cohesive pipeline stage.

ListCombinerNode

Splits text strings into a list of items using custom delimiters (supports commas, newlines, etc.) with optional empty value removal.

ListCreaterNode

Passes a list input through unchanged. Acts as a connector or placeholder in the graph to manage node connections without altering data content.

ListRerouteNode

Loads an image from the input directory, applies EXIF orientation correction, and outputs the image tensor along with an alpha mask (if present). Additionally, it provides both full raw metadata and a filtered 'clean' version containing only essential parameters for downstream workflows.

LoadImageWithMetadataNode

Aggregates up to 10 LoRA models with individually assigned weights, generating both a formatted prompt string and a structured list. Supports dynamic inclusion of empty values (placeholders) alongside active LoRAs for flexible pipeline construction.

LORASelectorNode

Inspects and reports the tensor shape of a connected mask node in string format. Useful for debugging pipeline issues related to dimension mismatches or verifying input consistency before further processing steps.

MaskDebugNode

Aggregates standard diffusion model weights from up to 10 inputs. Generates a combined list and comma-separated string of selected checkpoint files for flexible pipeline configuration.

ModelGeneratorNode

Creates a list of prompts, substituting all possible combinations of values from specified lists into placeholders in the template.

MultiPlaceholderPromptList

Buffers individual images into a visual grid as the XYZ loop progresses. Handles the accumulation of images per page, renders the preview grid with axis labels when full, and clears the batch upon completion of each slice.

MyXYGridAccumulator

Orchestrates the grid layout by mapping the current execution index to specific row, column, and page values. Dynamically generates annotations (e.g., 'Value: 5') for the XYZ plot headers based on input lists and styling parameters.

MyXYZHelper

Collects all rendered XY pages into a single multi-page image sequence once the full XYZ dataset is generated. Acts as the final aggregation node to output the complete result set for saving or further processing.

MyXYZSuperStacker

Merges 1 to 3 conditioning inputs (text embeddings) into a single output by averaging their weights and layering them element-wise. Automatically calculates the blend weight based on the number of active connections, ensuring seamless integration of multiple conditioning signals without manual scaling.

OptionalCondMergeNode

Retrieves patch model names from the ComfyUI patches folder and outputs both filename tuple (combo) and string representation for dynamic patch selection.

PatchModelSelectorNode

Dynamically constructs a prompt prefix string containing Pony-specific score, rating, and source tags (e.g., score_9, rating_safe). Combines selected variants into a single comma-separated string for immediate use in text encoding.

PonyPrefixesNode

An external controller panel for defining the resampling filter used by ImageResizeNode. Allows remote configuration of interpolation types (nearest, bilinear, bicubic, area, lanczos, etc.) by outputting the selected value to the main node's 'interpolation' input.

ResizeInterpolationControlNode

An external controller panel specifically for setting the resizing strategy of ImageResizeNode. Outputs the selected method (stretch, keep proportion, fill/crop, or pad) as a combo parameter to be connected directly to the main node's 'method' input.

ResizeMethodControlNode

Generates a comma-separated string and list of any selected samplers from up to 10 inputs. Useful for dynamically constructing sampler lists based on user selection before feeding them into sampling nodes.

SamplerGeneratorNode

Converts a string input into a valid KSampler sampler object. Automatically validates the input and falls back to 'euler' if an unsupported sampler name is provided, ensuring workflow stability.

SamplerSelectorFromStringNode

Saves images to the output folder without embedding workflow metadata. Supports PNG/JPEG formats, automatic date stamping (%date%) with custom format support, and auto-indexing (e.g., 00001.png) for duplicate files.

SaveImageNoMetaNode

Resizes an image to fit within a specified maximum dimension while strictly preserving its aspect ratio. Supports multiple interpolation modes (e.g., bilinear, bicubic, lanczos)

ScaleImageAspectNode

Generates a comma-separated string and list of any selected schedulers from up to 10 inputs. Used to dynamically assemble scheduler configurations from multiple source nodes before applying them to the sampling process.

SchedulerGeneratorNode

Parses a string input to select a compatible KSampler scheduler. Includes basic validation to handle incorrect inputs gracefully by defaulting to the 'normal' scheduler when necessary.

SchedulerSelectorFromStringNode

Applies a dynamic shift value to the sampling process or subsequent operations.

ShiftSliderNode

Converts a string value into a universal type-compatible output, allowing seamless bridging between different data types.

StringToAnyNode

Converts a text input string into a floating-point number with error handling. If the string cannot be parsed as a valid float, it logs the issue and returns 0.0 to keep the pipeline running smoothly.

StringToFloatNode

Safely converts a text input string into an integer. Automatically handles parsing errors by logging them and returning 0 as a fallback, ensuring the workflow continues without crashing on invalid numeric strings.

StringToIntNode

Merges up to 5 optional text inputs into a single continuous string, using a configurable delimiter.

TextConcatNode

Consolidates up to 10 selected text encoders (CLIP models) into a unified string output and list. Enables the dynamic assembly of multi-stage conditioning pipelines by combining various encoder variants.

TextEncoderGeneratorNode

Selects a single Text Encoder (CLIP model) from the installed encoders. Provides both the encoder path compatible with loading nodes and a corresponding string output, enabling flexible routing of conditioning models within the workflow.

TextEncoderSelectorNode

Aggregates up to 10 selected VAE models into a single comma-separated string and a corresponding list. Designed to dynamically build a chain of VAEs for processing workflows that require multiple decode/encode stages.

VAEGeneratorNode

Selects a single VAE model from the available list. Outputs both the model path (for direct connection to nodes like Load VAE) and a plain text string representation of the selection for use in other utility nodes.

VAESelectorNode

Output integer value with strict range constraints (min: 1, max: 10000, step: 4). 

WanNumFramesNode

Safeguards the XYZ pipeline against type mismatches by enforcing that only one active parameter exists per execution step. Automatically casts and outputs the current grid value (Row, Column, or Page) as an Int, Float, or String based on configuration.

ComfyUI Extension: DemonAlone-nodes-ComfyUI

Custom Nodes (53)

README

Credits:

Run ComfyUI workflows without the setup