ComfyUI-ppm

AttentionCouplePPM

> [!TIP] > > Connect this node after other model patches (such as CLIPNegPip, EasyCache, LazyCache, etc.). > > See "attentioncouple+negpip" and "attentioncouplewithglobalprompt" from ComfyUI's "Browse Templates" menu. > [!NOTE] > You can also use the version from ComfyUI prompt control (doc) - it has a convenient prompt-based approach that doesn't require to add new cond/mask nodes for each new region. Supports: - SD1 - SDXL - Anima Modified implementation of AttentionCouple by laksjdjf and Haoming02, made to be more compatible with other custom nodes. Inputs for new regions are managed automatically: when you attach cond/mask of a region to the node, a new cond / mask input appears. Link basecond input to the positive conditioning used in KSampler/SamplerCustom. You can use multiple LatentToMaskBB nodes to set bounding boxes for AttentionCouplePPM. The parameters are relative to your initial latent: x=0.5, y=0.0, w=0.5, h=1.0 will produce a mask covering the right half of the image. You can adjust mask values to set region strength and use ConditioningSetAreaStrength to increase/decrease conditioning strength.

3 inputs · 0 parameters

CADSPPM

Clones the model and patches its conditioning calculation to add noise to model-conditioning tensors within the configured sigma interval, derived from start_percent and end_percent; scale controls the noise amount.

1 input · 3 parameters

CFGLimiterGuider

Builds a CFG-limiter guider from the model, positive and negative conditioning, CFG value, and sigma limits.

Sampling/Custom Sampling/Guiders · 1 output

CFGPPSamplerSelect

Creates a sampler from the selected CFG++ sampler; supplies eta for eta samplers and gamma interval and extra-step options for dynamic CFG++ samplers.

Model/Sampling/Samplers · 1 output

CLIPAttentionSelector

Clones the CLIP object and replaces the text model's optimized attention override with the selected registered attention function.

1 input · 1 parameter

CLIPMicroConditioning

Sets width, height, crop width, crop height, target width, and target height values on the supplied conditioning.

Advanced/Conditioning · 1 output

CLIPNegPip

> [!TIP] > > See "attentioncouple+negpip" or "fluxnegpip" from ComfyUI's "Browse Templates" menu. Allows you to use negative weights in prompts. Negative weights can be used in both CFG models (such as SDXL) and Guidance-Distilled models (such as Flux) to negate a concept or a trait. Supports: - SD1 - SDXL - Anima - FLUX Unmaintained Modified implementation of NegPiP by laksjdjf and hako-mikan. It uses ModelPatcher instead of monkey-patching, which should increase compatibility with other nodes. You can read more about NegPiP in the original repo. When used together with tag-based models, you should keep all commas inside weight braces (i.e. (worst quality,:-1.3) (sketch,:-1.1) instead of (worst quality:-1.3), (sketch:-1.1),). Some tag-based models (Illustrious/NoobAI) perform better with commas in braces, while other models (like Anima) perform better with commas outside of emphasis blocks - feel free to experiment. > [!NOTE] > CLIPNegPip is compatible with: > > - ComfyUI prompt control by asagi4 > - Advanced CLIP Text Encode extension by BlenderNeko > > CLIPNegPip is incompatible with: > > - smZ Nodes by shiimizu (ComfyUI prompt control can replace most of its functionality) > - ComfyuiFluxStyleAdjust by yichengup (and probably some other custom nodes that modify cond tensors)

2 inputs · 0 parameters

CLIPTextEncodeBREAK

A small lightweight wrapper over ConditioningConcat node. It splits prompts into chunks by BREAK keyword and produces a single concatenated conditioning.

Conditioning · 1 output

CLIPTextEncodeInvertWeights

Splits text at BREAK, encodes each chunk with token weights inverted, preserves special-token weights unless invert_special_tokens is enabled, and concatenates the resulting conditioning.

Conditioning · 1 output

CLIPTokenCounter

Counts tokens in your prompt and returns them as a string. You can use Preview Any node to display them.

3 outputs

ConditioningZeroOutCombine

Combines zeroed conditioning over the timestep range from 0 to zero_out_end with the original conditioning over the range from zero_out_end to 1.

Advanced/Conditioning · 1 output

ConvertTimestepToSigma

Converts a model sampling percentage to sigma, optionally returning the actual sigma at 0 or 1, or returns the sigma at a selected schedule step.

1 input · 1 parameter

DynamicThresholdingPost

Adds a post-CFG function that applies dynamic thresholding to the CFG denoised output using mimic-scale guidance and the configured thresholding options.

Model Patches/Unet · 1 output

DynamicThresholdingSimplePost

Applies DynamicThresholdingPost with its default optional thresholding settings, exposing the model, mimic scale, and threshold percentile inputs.

Model Patches/Unet · 1 output

DynSamplerSelect

Modified samplers from Euler-Smea-Dyn-Sampler by Koishi-Star. Contains some new samplers: eulerancestraldy, dpmpp2mdy and dpmpp3mdy. Tweaking sdypow may reduce blur artifacts (optimal value is 2 for euler samplers and -1 for dpmpp samplers, use -1 to disable this feature).

Sampling/Custom Sampling/Samplers · 1 output

EmptyLatentImageAR

Creates an empty latent batch with dimensions calculated from the resolution, aspect ratio, and step inputs.

Latent · 1 output

EpsilonScalingPPM

Modified version of ComfyUI's Epsilon Scaling node with a hacky (and mathematically incorrect) support for v-pred ZSNR models.

1 input · 1 parameter

FreeU2PPM

Compact description unavailable for this node.

For Testing · 1 output

Guidance Limiter

Implementation of Applying Guidance in a Limited Interval Improves Sample and Distribution Quality in Diffusion Models by Kynkäänniemi et al. as a post CFG function. Guidance Limiter is also available as a CFGLimiterGuider guider node for SamplerCustomAdvanced.

Model Patches/Unet · 1 output

LatentOperationTonemapLuminance

Compact description unavailable for this node.

Latent/Advanced/Operations · 1 output

LatentToMaskBB

Sets bounding-box masks for AttentionCouplePPM using coordinates relative to the initial latent.

Mask · 1 output

LatentToWidthHeight

Compact description unavailable for this node.

Latent · 2 outputs

MaskCompositePPM

Compact description unavailable for this node.

Mask · 1 output

ModelAttentionSelector

Can swap the optimized attention algorithm, including pytorch SDPA, sage, and xformers, without restarting ComfyUI.

1 input · 1 parameter

PPMSamplerSelect

Compact description unavailable for this node.

Model/Sampling/Samplers · 1 output

RenormCFGPost

Post-CFG variant of RenormCFG; the README says these post-CFG variants can be chained with other CFG-related nodes for increased compatibility.

Model Patches/Unet · 1 output

RescaleCFGPost

Post-CFG variant of RescaleCFG; the README says these post-CFG variants can be chained with other CFG-related nodes for increased compatibility.

Model Patches/Unet · 1 output

SamplerER_SDEScheduled

Compact description unavailable for this node.

1 input · 6 parameters

SamplerGradientEstimation

Compact description unavailable for this node.

Model/Sampling/Samplers · 1 output

SamplerSEEDS2Scheduled

Compact description unavailable for this node.

1 input · 6 parameters

SkipFirstStepCFG

Compact description unavailable for this node.

Model Patches/Unet · 1 output

TCFGAdvanced

TCFG – Tangential Damping CFG (2503.18137)\n\nRefine the uncond (negative) to align with the cond (positive) for improving quality.

Advanced/Guidance · 1 output

TilePreprocessorPPM

Image preprocessor for ControlNet Tile that doesn't require any third-party libraries (aside from kornia, which is a part of ComfyUI's requirements).

1 input · 4 parameters