Skoogeer-Noise
A noise node pack for ComfyUI enabling all kinds of crazy noise to be added to latents, conditioning vectors, and images.
A noise node pack for ComfyUI enabling all kinds of crazy noise to be added to latents, conditioning vectors, and images.
Paketprofil
- Status
- Aufgelistet
- Verleger
- @ttulttul
- Quelle
- GitHub
- Sterne
- 14
README
A noise node pack for ComfyUI enabling all kinds of crazy noise to be added to latents, conditioning vectors, and images.
Knoten in diesem Paket
63 Knoten (Plural)
Knoten in diesem Paket
63 Knoten (Plural)- AnythingToMarkdown
AnythingToMarkdown
Formats any connected value as a readable Markdown summary. It detects common ComfyUI-native structures such as LATENT, IMAGE, and CONDITIONING, and also reports regular Python container and scalar types. - Menu category: text/debug - Returns: STRING | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | anything | any | – | – | Value to inspect. Lists are inspected item-by-item. | | Output | Type | Description | |------|------|-------------| | markdown | STRING | Markdown text describing the detected type and contents. | - The node publishes the same Markdown through its own ComfyUI text preview payload, so it can be used directly as an output node. - The markdown socket is emitted as a single list-aware STRING item so downstream text nodes receive the Markdown document itself rather than a Python list wrapper. - Tensors are summarized by shape, dtype, device, element count, min, max, mean, and standard deviation. - Tensor contents are not dumped into the output. - Container summaries are bounded so large inputs remain previewable.
Text/Debug · 1 Eingabe · 0 Parameter · 1 Ausgabe
- CombineLatentPhaseMagnitude
CombineLatentPhaseMagnitude
Converts magnitude + phase frequency-domain latents back into a spatial latent.
Latent/Frequency · 2 Eingaben · 0 Parameter · 1 Ausgabe
- ConcatenateLists
ConcatenateLists
Concatenates two list-valued inputs into one list without changing the item type. This is useful when you want to append batches, not when you need to combine per-entry dictionaries. - Menu category: utils/list - Returns: list of the same item type | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | listitems1 | list | – | – | First list-valued input. | | listitems2 | list | – | – | Second list-valued input, appended after listitems1. | | Output | Type | Description | |------|------|-------------| | listitems | list | Concatenation of listitems1 followed by listitems2. | - This node is generic and works with MUSTACHEVARIABLELIST, STRING lists, and other list-valued sockets. - For the mustache workflow, the list-valued node is Mustache Variable Sampler; Mustache Variables itself still returns a MUSTACHEVARIABLES mapping. - For mustache workflows, use Merge Mustache Variable Sets instead when a downstream template needs each output entry to include keys from both input branches.
Utils/List · 2 Eingaben · 0 Parameter · 1 Ausgabe
- ConditioningAddNoise
ConditioningAddNoise
Adds seeded Gaussian noise to conditioning embeddings and their pooledoutput when present. - Menu category: conditioning/noise - Returns: CONDITIONING | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | conditioning | CONDITIONING | – | – | Conditioning list to perturb with Gaussian noise. | | seed | INT | 0 | 0..2^64-1 | Seed that drives the conditioning noise. | | strength | FLOAT | 0.1 | 0.0..5.0 | Noise strength relative to each tensor's standard deviation. |
Conditioning/Noise · 1 Eingabe · 2 Parameter · 1 Ausgabe
- ConditioningFrequencyMerge
ConditioningFrequencyMerge
Recombines low/high conditioning bands back into a single conditioning list. - Menu category: conditioning/filter - Returns: CONDITIONING | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | lowpass | CONDITIONING | – | – | Low-frequency conditioning list produced by the split node. | | highpass | CONDITIONING | – | – | High-frequency conditioning list to recombine. | | lowgain | FLOAT | 1.0 | -5.0..5.0 | Multiplier for the low-pass band before merging. | | highgain | FLOAT | 1.0 | -5.0..5.0 | Multiplier for the high-pass band before merging. |
Conditioning/Filter · 2 Eingaben · 2 Parameter · 1 Ausgabe
- ConditioningFrequencySplit
ConditioningFrequencySplit
Separates conditioning embeddings into low/high bands via Gaussian smoothing along the token axis. - Menu category: conditioning/filter - Returns: CONDITIONING (low), CONDITIONING (high) | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | conditioning | CONDITIONING | – | – | Conditioning list to separate into low/high bands. | | sigma | FLOAT | 0.75 | 0.0..10.0 | Cutoff for the Gaussian low-pass applied along the token axis. |
Conditioning/Filter · 1 Eingabe · 1 Parameter · 2 Ausgaben
- ConditioningGaussianBlur
ConditioningGaussianBlur
Applies Gaussian smoothing along the token dimension of conditioning embeddings. - Menu category: conditioning/filter - Returns: CONDITIONING | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | conditioning | CONDITIONING | – | – | Conditioning list whose token dimension will be blurred. | | sigma | FLOAT | 0.75 | 0.0..10.0 | Standard deviation of the blur kernel along the token axis. |
Conditioning/Filter · 1 Eingabe · 1 Parameter · 1 Ausgabe
- ConditioningScale
ConditioningScale
Scales conditioning embeddings (and pooledoutput when present) to amplify or mute prompt influence. - Menu category: conditioning/filter - Returns: CONDITIONING | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | conditioning | CONDITIONING | – | – | Conditioning list to scale. | | factor | FLOAT | 1.0 | 0.0..10.0 | Multiplier applied to embeddings. 0 mutes, 1 keeps original. |
Conditioning/Filter · 1 Eingabe · 1 Parameter · 1 Ausgabe
- FluidImageAdvection
FluidImageAdvection
Distorts an image by advecting pixels through a viscous velocity field and returns a velocity preview image. - Menu category: image/perturb - Returns: IMAGE, IMAGE See docs/fluid-simulation.md for the full parameter list (shared with the latent variant). Core controls include steps, dt, resolutionscale, force injection settings, vorticity, and wrapmode. - Optional mask limits the effect to masked areas.
Image/Perturb · 2 Eingaben · 12 Parameter · 2 Ausgaben
- FluidLatentAdvection
FluidLatentAdvection
Advects latent channels through a viscous velocity field and returns a velocity preview image. - Menu category: latent/perturb - Returns: LATENT, IMAGE See docs/fluid-simulation.md for the full parameter list (shared with the image variant). Core controls include steps, dt, resolutionscale, force injection settings, vorticity, and wrapmode. - Optional mask limits the effect to masked areas.
Latent/Perturb · 2 Eingaben · 12 Parameter · 2 Ausgaben
- FrequencySelectiveStructuredNoise
FrequencySelectiveStructuredNoise
Generates frequency-domain structured noise using a reference phase latent. The output magnitude is taken entirely from random Gaussian noise (FFT magnitude). The output phase is mixed between the reference phase and noise phase using a radial frequency mask.
Latent/Frequency · 1 Eingabe · 3 Parameter · 2 Ausgaben
- ImageAddNoise
ImageAddNoise
Adds seeded Gaussian noise to image tensors.
Image/Noise · 2 Eingaben · 2 Parameter · 1 Ausgabe
- ImageBatchToLatent
ImageBatchToLatent
Merges a batch of per-channel grayscale images back into a LATENT tensor. - Menu category: latent/debug - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | imagebatch | IMAGE | – | – | Batch of images representing latent channels. | | batchsize | INT | 0 | 0..4096 | Original latent batch size (0 to infer from channels). | | channels | INT | 0 | 0..4096 | Channel count per latent sample (0 to infer from batchsize). | | channelsource | enum | r | r/g/b/mean | Which channel to use from RGB inputs (mean averages channels). |
Latent/Debug · 1 Eingabe · 3 Parameter · 1 Ausgabe
- ImageFractalBrownianMotion
ImageFractalBrownianMotion
Builds fractal Brownian motion (fBm) from a selectable base noise and injects it into an image. - Menu category: Image/Noise - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image to enrich with fractal Brownian motion. | | seed | INT | 0 | 0..2^64-1 | Seed for the base noise generator. | | basenoise | enum | simplex | simplex/perlin/worley | Noise primitive accumulated by the fBm stack. | | frequency | FLOAT | 2.0 | 0.01..64.0 | Fundamental frequency for simplex/perlin (acts as a multiplier for Worley). | | featurepoints | INT | 16 | 1..4096 | Base feature point count (used when basenoise = worley). | | octaves | INT | 5 | 1..12 | Number of fBm layers to accumulate. | | persistence | FLOAT | 0.5 | 0.0..1.0 | Amplitude multiplier between fBm layers. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Frequency multiplier between fBm layers. | | distancemetric | enum | euclidean | euclidean/manhattan/chebyshev | Distance metric used when the base noise is Worley. | | jitter | FLOAT | 0.35 | 0.0..1.0 | Feature jitter amount for Worley base noise. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized fBm relative to the image's standard deviation. | | channelmode | enum | shared | shared/perchannel | Shared fBm field per sample or reseeded per channel. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Image/Noise · 2 Eingaben · 12 Parameter · 1 Ausgabe
- ImageMeshDrag
ImageMeshDrag
Applies the same mesh-drag deformation in image space (pixel units) to ComfyUI IMAGE tensors. - Menu category: image/perturb - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image to warp spatially using a random mesh drag. | | seed | INT | 0 | 0..2^64-1 | Seed controlling which mesh points are dragged and by how much. | | points | INT | 12 | 0..4096 | Number of mesh vertices to randomly drag. | | dragmin | FLOAT | 0.0 | 0.0..4096.0 | Minimum drag distance (image pixels). | | dragmax | FLOAT | 32.0 | 0.0..4096.0 | Maximum drag distance (image pixels). | | direction | FLOAT | -1.0 | -1.0..360.0 | Drag direction in degrees (0=up, 90=right, 180=down, 270=left). -1 allows random directions. | | strokewidth | FLOAT | -1.0 | -1.0..16384.0 | When >0, limits the warp to narrow brush-stroke channels aligned with the direction (one per dragged mesh point; when direction>=0, stroke centers are jittered so they aren't grid-aligned). Units are image pixels. -1 disables. | | displacementinterpolation | enum | bicubic | bilinear/bicubic/bspline/nearest | How to interpolate sparse mesh drags into a full displacement field (bspline is smoother). | | splinepasses | INT | 2 | 0..16 | Only used when displacementinterpolation = bspline. | | samplinginterpolation | enum | bilinear | bilinear/bicubic/nearest | How to sample the source image when applying the warp. |
Image/Perturb · 1 Ausgabe
- ImageNoise
ImageNoise
Adds seeded Gaussian noise to a ComfyUI IMAGE tensor. - Menu category: image/perturb - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image to receive additional Gaussian noise. | | seed | INT | 0 | 0..2^64-1 | See Common Parameters. | | strength | FLOAT | 1.0 | 0.0..10.0 | See Common Parameters. |
Image/Perturb · 2 Eingaben · 2 Parameter · 1 Ausgabe
- ImagePerlinFractalNoise
ImagePerlinFractalNoise
Adds smooth fractal Perlin noise to an image for structured variation. - Menu category: Image/Noise - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image that will be perturbed with fractal Perlin noise. | | seed | INT | 0 | 0..2^64-1 | Seed controlling the procedural noise pattern. | | frequency | FLOAT | 2.0 | 0.01..64.0 | Base lattice frequency (higher = finer detail). | | octaves | INT | 4 | 1..12 | Number of noise layers to accumulate. | | persistence | FLOAT | 0.5 | 0.0..1.0 | Amplitude multiplier between octaves. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Frequency multiplier between octaves. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized noise relative to the image's standard deviation. | | channelmode | enum | shared | shared/perchannel | See Common Parameters. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Image/Noise · 2 Eingaben · 8 Parameter · 1 Ausgabe
- ImageReactionDiffusion
ImageReactionDiffusion
Runs a Gray-Scott reaction-diffusion simulation and injects the resulting pattern into an image. - Menu category: Image/Noise - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image that will receive reaction-diffusion patterns. | | seed | INT | 0 | 0..2^64-1 | Seed for the initial chemical concentrations. | | iterations | INT | 200 | 1..2000 | Number of Gray-Scott simulation steps. | | feedrate | FLOAT | 0.036 | 0.0..0.1 | Feed rate (F) controlling how quickly chemical U is replenished. | | killrate | FLOAT | 0.065 | 0.0..0.1 | Kill rate (K) regulating removal of chemical V. | | diffusionu | FLOAT | 0.16 | 0.0..1.0 | Diffusion rate for chemical U. | | diffusionv | FLOAT | 0.08 | 0.0..1.0 | Diffusion rate for chemical V. | | timestep | FLOAT | 1.0 | 0.01..5.0 | Simulation time step used during integration. | | strength | FLOAT | 0.75 | 0.0..5.0 | Scales normalized pattern relative to the image's standard deviation. | | channelmode | enum | shared | shared/perchannel | Reuse one simulation for all channels or rerun per channel. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Image/Noise · 2 Eingaben · 10 Parameter · 1 Ausgabe
- ImageSimplexNoise
ImageSimplexNoise
Adds layered fractal simplex noise to an image. - Menu category: Image/Noise - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image to perturb with simplex noise. | | seed | INT | 0 | 0..2^64-1 | Seed controlling the simplex lattice offsets. | | frequency | FLOAT | 2.0 | 0.01..64.0 | Base lattice frequency for the simplex grid. | | octaves | INT | 4 | 1..12 | Number of simplex layers to accumulate. | | persistence | FLOAT | 0.5 | 0.0..1.0 | Amplitude multiplier applied between octaves. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Frequency multiplier applied between octaves. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized simplex noise relative to the image's standard deviation. | | channelmode | enum | shared | shared/perchannel | See Common Parameters. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Image/Noise · 2 Eingaben · 8 Parameter · 1 Ausgabe
- ImageSmokeSimulation
ImageSmokeSimulation
Simulates buoyancy-driven smoke in image space and returns density and velocity previews. - Menu category: image/perturb - Returns: IMAGE, IMAGE, IMAGE See docs/fluid-simulation.md for smoke-mode parameters such as smokesourcemode, buoyancy, cooling, and output batching controls.
Image/Perturb · 2 Eingaben · 21 Parameter · 3 Ausgaben
- ImageSwirlNoise
ImageSwirlNoise
Swirls image pixels around randomized centers (vortex-like warp). This is a spatial deformation, not additive noise. - Menu category: Image/Noise - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image to deform with vortex-style warps. | | seed | INT | 0 | 0..2^64-1 | Seed for vortex placement and direction randomness. | | vortices | INT | 1 | 1..16 | Number of independent vortex centres to spawn per image. | | channelmode | enum | global | global/perchannel | Shared swirl grid for all affected channels or unique grid per channel. | | channelfraction | FLOAT | 1.0 | 0.0..1.0 | Fraction of channels to swirl (subset chosen per sample). | | strength | FLOAT | 0.75 | 0.0..6.28 | Peak swirl rotation in radians near the vortex center. | | radius | FLOAT | 0.5 | 0.05..2.0 | Normalized radius controlling how far the vortex influence extends. | | centerspread | FLOAT | 0.25 | 0.0..1.0 | How far the vortex origin drifts from the image center. | | directionbias | FLOAT | 0.0 | -1.0..1.0 | Bias toward counter-clockwise (1) or clockwise (-1) swirl. | | mix | FLOAT | 1.0 | 0.0..1.0 | See Common Parameters. |
Image/Noise · 2 Eingaben · 9 Parameter · 1 Ausgabe
- ImageToBatch
ImageToBatch
Replaces or inserts an image or image batch in an existing IMAGE batch at a selected batch index. - Menu category: image/batch - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | imagebatch | IMAGE | - | - | Existing image batch to modify. | | image | IMAGE | - | - | Image or image batch to place. Spatial shape, channels, dtype, and device must match imagebatch. | | index | INT | 0 | 0..4096 | Target batch index. In replace mode it must point at an existing item; in insert mode it can also equal the current batch size to append. | | mode | enum | replace | replace/insert | replace overwrites existing item(s). insert shifts existing items right. |
Image/Batch · 2 Eingaben · 2 Parameter · 1 Ausgabe
- ImageWorleyNoise
ImageWorleyNoise
Generates cellular (Worley) noise and injects it into an image. - Menu category: Image/Noise - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | image | IMAGE | – | – | Image to perturb with Worley (cellular) noise. | | seed | INT | 0 | 0..2^64-1 | Seed for the feature point distribution. | | featurepoints | INT | 16 | 1..4096 | Base number of feature points scattered across the plane. | | octaves | INT | 3 | 1..8 | Number of cellular layers to accumulate. | | persistence | FLOAT | 0.65 | 0.0..1.0 | Amplitude multiplier between Worley octaves. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Multiplier for the feature point count between octaves. | | distancemetric | enum | euclidean | euclidean/manhattan/chebyshev | Distance metric used when measuring feature proximity. | | jitter | FLOAT | 0.35 | 0.0..1.0 | How far feature points can drift inside each cell. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized Worley noise relative to the image's standard deviation. | | channelmode | enum | shared | shared/perchannel | See Common Parameters. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Image/Noise · 2 Eingaben · 10 Parameter · 1 Ausgabe
- JoinTextList
JoinTextList
Consumes a list-valued STRING input and joins every item into one string so preview/debug nodes can show the whole batch at once. - Menu category: text/debug - Returns: STRING, INT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | textitems | STRING list | – | – | List-valued string input, such as the output of Mustache Template. | | separator | STRING | \n | multiline | Text inserted between each item. Escape sequences like \n, \r, and \t are decoded, so values like \n===\n produce visible section breaks. | | Output | Type | Description | |------|------|-------------| | joinedtext | STRING | All list items joined with the requested separator. | | count | INT | Number of input strings that were joined. | - The node unwraps singleton-list widget values for separator, which matches how ComfyUI sometimes passes control inputs into list-aware nodes. text Mustache Template -> Join Text List -> Preview as Text
Text/Debug · 0 Eingaben · 2 Parameter · 2 Ausgaben
- KSamplerLoraSigmaInverse
KSamplerLoraSigmaInverse
KSampler-compatible sampler node with an embedded model-only LoRA loader. The LoRA strength is changed per sigma step using: strength[i] = minlorastrength + (maxlorastrength - minlorastrength) (1 - sigma[i] / max(sigma)) This means LoRA influence starts at minlorastrength at high sigma and moves toward maxlorastrength as sigma decreases. You can also gate LoRA application by step index: - If i < minlorastep, effective strength is 0. - If i > maxlorastep, effective strength is 0. - -1 for either bound disables that bound (always enabled on that side). - Menu category: sampling - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | model | MODEL | – | – | Diffusion model to denoise with. | | seed | INT | 0 | 0..2^64-1 | Seed for latent noise preparation. | | steps | INT | 20 | 1..10000 | Sampling steps. | | cfg | FLOAT | 8.0 | 0.0..100.0 | CFG scale. | | samplername | enum | euler | ComfyUI sampler list | Sampling algorithm. | | scheduler | enum | normal | ComfyUI scheduler list | Sigma schedule family. | | positive | CONDITIONING | – | – | Positive conditioning. | | negative | CONDITIONING | – | – | Negative conditioning. | | latentimage | LATENT | – | – | Input latent. | | loraname | enum | first available | LoRA files in loras | LoRA to schedule. | | minlorastrength | FLOAT | 0.0 | -100.0..100.0 | Starting LoRA strength at maximum sigma. | | maxlorastrength | FLOAT | 1.0 | -100.0..100.0 | Final LoRA strength at minimum sigma. | | minlorastep | INT | -1 | -1..10000 | First enabled step index; -1 = no lower bou…
Sampling · 1 Ausgabe
- LatentAddNoise
LatentAddNoise
Adds a configurable amount of seeded random noise to a latent tensor. The strength is relative to the standard deviation of the input latent.
Latent/Noise · 2 Eingaben · 2 Parameter · 1 Ausgabe
- LatentChannelLinearTransform
LatentChannelLinearTransform
Applies linear channel-space transforms (signed permutations, orthogonal rotations, Householder reflections, low-rank shears). - Menu category: latent/channel - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to transform in channel-space. | | operation | enum | signedpermute | signedpermute/orthogonalrotate/householderreflect/lowrankshear | Linear transform to apply. | | seed | INT | 0 | 0..2^64-1 | Seed for deterministic selection/transforms. | | signflipprob | FLOAT | 0.5 | 0.0..1.0 | Sign-flip probability for signed permutations. | | tilesize | INT | 0 | 0..512 | Tile size (latent pixels) for per-tile permutations. | | blocksize | INT | 0 | 0..4096 | Channels per orthogonal block (0 = full). | | alpha | FLOAT | 0.5 | -4.0..4.0 | Shear strength for low-rank shear. | | selectionmode | enum | all | all/random/topvariance/toproughness/indices | Channel selection strategy. | | selectionfraction | FLOAT | 1.0 | 0.0..1.0 | Fraction of channels when selectioncount=0. | | selectioncount | INT | 0 | 0..4096 | Exact number of channels (overrides fraction). | | selectionorder | enum | highest | highest/lowest | Choose high or low variance/roughness. | | selectionindices | STRING | "" | – | Comma-separated indices when selectionmode=indices. | | mix | FLOAT | 1.0 | 0.0..1.0 | See Common Parameters. | | matchstats | BOOLEAN | false | – | Match per-channel mean/std after edit. | | mask | MASK | – | – | Optional mask to limit the transform. | | Mode | What it does | Likely impact | |-…
Latent/Channel · 1 Ausgabe
- LatentChannelMatch
LatentChannelMatch
Transfers the per-channel mean and standard deviation from a reference latent onto a target latent, similar to a color-match operation but performed directly in latent space. - Menu category: latent/channel - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | target | LATENT | – | – | Latent whose channels will be normalized to match the reference statistics. | | reference | LATENT | – | – | Latent providing the per-channel mean/std to transfer. | | mix | FLOAT | 1.0 | 0.0..1.0 | Blend factor between the original target and the fully matched result. | | mask | MASK | – | – | Optional mask to limit the matched result spatially. | - Matching is done independently per channel using the reference channel mean/std and the target channel mean/std. - The reference latent can either flatten to one sample or to the same flattened sample count as the target. A single reference sample broadcasts across the target batch. - Spatial sizes may differ between reference and target because only channel statistics are transferred. - This is useful for latent-space “palette transfer” or for normalizing one latent toward another latent’s overall channel distribution before downstream blending or decoding.
Latent/Channel · 3 Eingaben · 1 Parameter · 1 Ausgabe
- LatentChannelMerge
LatentChannelMerge
Blends selected channels from a source latent into a destination latent. - Menu category: latent/channel - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | destination | LATENT | – | – | Latent to blend into (destination). | | source | LATENT | – | – | Latent providing channels to blend (source). | | seed | INT | 0 | 0..2^64-1 | Seed for deterministic selection when selectionmode=random. | | selectionmode | enum | all | all/random/topvariance/toproughness/indices | Channel selection strategy (metrics measured on the source latent). | | selectionfraction | FLOAT | 1.0 | 0.0..1.0 | Fraction of channels when selectioncount=0. | | selectioncount | INT | 0 | 0..4096 | Exact number of channels (overrides fraction). | | selectionorder | enum | highest | highest/lowest | Choose high or low variance/roughness. | | selectionindices | STRING | "" | – | Comma-separated indices when selectionmode=indices. | | blendstrength | FLOAT | 1.0 | -4.0..4.0 | Blend factor for selected channels (0 keeps destination, 1 uses source, values outside [0,1] extrapolate). | | mask | MASK | – | – | Optional mask to limit the merge. |
Latent/Channel · 3 Eingaben · 7 Parameter · 1 Ausgabe
- LatentChannelNonlinearTransform
LatentChannelNonlinearTransform
Applies nonlinear channel-space transforms (gating, quantization, clipping, dropout/replacement). - Menu category: latent/channel - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to transform in channel-space. | | operation | enum | gatemultiply | gatemultiply/gateadd/quantize/cliphard/clipsoft/dropoutzero/dropoutnoise/dropoutswap | Nonlinear transform to apply. | | seed | INT | 0 | 0..2^64-1 | Seed for deterministic gating/dropout. | | gatestrength | FLOAT | 1.0 | 0.0..10.0 | Gate slope (gate operations). | | beta | FLOAT | 1.0 | -4.0..4.0 | Gate scale (gate operations). | | blurradius | INT | 0 | 0..64 | Blur radius for gate maps (latent pixels). | | quantizestep | FLOAT | 0.25 | 0.0..10.0 | Step size for quantization. | | clipthreshold | FLOAT | 2.0 | 0.0..10.0 | Threshold for clipping. | | selectionmode | enum | all | all/random/topvariance/toproughness/indices | Channel selection strategy. | | selectionfraction | FLOAT | 1.0 | 0.0..1.0 | Fraction of channels when selectioncount=0. | | selectioncount | INT | 0 | 0..4096 | Exact number of channels (overrides fraction). | | selectionorder | enum | highest | highest/lowest | Choose high or low variance/roughness. | | selectionindices | STRING | "" | – | Comma-separated indices when selectionmode=indices. | | mix | FLOAT | 1.0 | 0.0..1.0 | See Common Parameters. | | matchstats | BOOLEAN | false | – | Match per-channel mean/std after edit. | | mask | MASK | – | – | Optional mask to limit the transform. | | Mode | What it d…
Latent/Channel · 1 Ausgabe
- LatentChannelStatsPreview
LatentChannelStatsPreview
Renders a bar-chart preview of per-channel mean and standard deviation for LATENT["samples"]. - Menu category: latent/debug - Returns: IMAGE | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to analyze. | | channellimit | INT | 16 | 1..64 | Number of channels to display. | | height | INT | 256 | 72..1024 | Output image height (layout adjusts automatically). |
Latent/Debug · 1 Eingabe · 2 Parameter · 1 Ausgabe
- LatentForwardDiffusion
LatentForwardDiffusion
Applies the 'natural' forward diffusion process to a clean latent. This produces a statistically 'perfect' noisy latent that samplers expect.
Latent/Noise · 3 Eingaben · 3 Parameter · 1 Ausgabe
- LatentFractalBrownianMotion
LatentFractalBrownianMotion
Builds fractal Brownian motion (fBm) from a selectable base noise and injects it into a latent. - Menu category: Latent/Noise - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to enrich with fractal Brownian motion. | | seed | INT | 0 | 0..2^64-1 | Seed for the base noise generator. | | basenoise | enum | simplex | simplex/perlin/worley | Noise primitive accumulated by the fBm stack. | | frequency | FLOAT | 2.0 | 0.01..64.0 | Fundamental frequency for simplex/perlin (acts as a multiplier for Worley). | | featurepoints | INT | 16 | 1..4096 | Base feature point count (used when basenoise = worley). | | octaves | INT | 5 | 1..12 | Number of fBm layers to accumulate. | | persistence | FLOAT | 0.5 | 0.0..1.0 | Amplitude multiplier between fBm layers. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Frequency multiplier between fBm layers. | | distancemetric | enum | euclidean | euclidean/manhattan/chebyshev | Distance metric used when the base noise is Worley. | | jitter | FLOAT | 0.35 | 0.0..1.0 | Feature jitter amount for Worley base noise. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized fBm relative to the latent's standard deviation. | | channelmode | enum | shared | shared/perchannel | Shared fBm field per sample or reseeded per channel. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Latent/Noise · 2 Eingaben · 12 Parameter · 1 Ausgabe
- LatentFrequencyMerge
LatentFrequencyMerge
Merges the low-pass and high-pass bands back into a single latent. - Menu category: Latent/Filter - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | lowpass | LATENT | – | – | Low-frequency latent band produced by Latent Frequency Split. | | highpass | LATENT | – | – | High-frequency latent band to merge back in. | | lowgain | FLOAT | 1.0 | -5.0..5.0 | Multiplier for the low-pass band before merging. | | highgain | FLOAT | 1.0 | -5.0..5.0 | Multiplier for the high-pass band before merging. | - merged: lowpass lowgain + highpass highgain
Latent/Filter · 3 Eingaben · 2 Parameter · 1 Ausgabe
- LatentFrequencySplit
LatentFrequencySplit
Splits a latent into low-pass and high-pass bands by subtracting a Gaussian-smoothed version. - Menu category: Latent/Filter - Returns: LATENT (low), LATENT (high) | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to decompose into low/high frequency bands. | | sigma | FLOAT | 1.5 | 0.0..20.0 | Radius of Gaussian low-pass. Higher moves more detail into the high band. | - lowpass: blurred latent - highpass: latent - lowpass (zeros when sigma <= 0)
Latent/Filter · 2 Eingaben · 1 Parameter · 2 Ausgaben
- LatentGaussianBlur
LatentGaussianBlur
Gaussian blur for latents (supports both 4D and 5D latents). - Menu category: Latent/Filter - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent tensor to blur. | | sigma | FLOAT | 1.0 | 0.0..10.0 | Standard deviation for the Gaussian kernel. 0 is a no-op. | | blurmode | enum | Spatial Only | Spatial Only/Spatial and Channel | When Spatial and Channel, also blurs across the channel dimension. |
Latent/Filter · 2 Eingaben · 2 Parameter · 1 Ausgabe
- LatentMeshDrag
LatentMeshDrag
Applies a cloth-like spatial warp to a LATENT by randomly dragging control vertices on a coarse mesh and interpolating a smooth displacement field. - Menu category: latent/perturb - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to warp spatially using a random mesh drag. | | mask | MASK | – | – | Optional mask to limit the warp to masked areas. The mask is resized to latent resolution (bicubic when downscaling). | | seed | INT | 0 | 0..2^64-1 | Seed controlling which mesh points are dragged and by how much. | | points | INT | 12 | 0..2048 | Number of mesh vertices to randomly drag. | | dragmin | FLOAT | 0.0 | 0.0..128.0 | Minimum drag distance (latent pixels). | | dragmax | FLOAT | 4.0 | 0.0..128.0 | Maximum drag distance (latent pixels). | | direction | FLOAT | -1.0 | -1.0..360.0 | Drag direction in degrees (0=up, 90=right, 180=down, 270=left). -1 allows random directions. | | strokewidth | FLOAT | -1.0 | -1.0..4096.0 | When >0, limits the warp to narrow brush-stroke channels aligned with the direction (one per dragged mesh point; when direction>=0, stroke centers are jittered so they aren't grid-aligned). Units are latent pixels. -1 disables. | | displacementinterpolation | enum | bicubic | bilinear/bicubic/bspline/nearest | How to interpolate sparse mesh drags into a full displacement field (bspline is smoother). | | splinepasses | INT | 2 | 0..16 | Only used when displacementinterpolation = bspline. | | samplinginterpolation | enum | bilinear | bilinear/bicubic/neare…
Latent/Perturb · 1 Ausgabe
- LatentNoise
LatentNoise
Adds seeded Gaussian noise to a ComfyUI LATENT dictionary. - Menu category: latent/perturb - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to receive additional Gaussian noise. | | seed | INT | 0 | 0..2^64-1 | See Common Parameters. | | strength | FLOAT | 1.0 | 0.0..10.0 | See Common Parameters. |
Latent/Perturb · 2 Eingaben · 2 Parameter · 1 Ausgabe
- LatentPackedSlotTransform
LatentPackedSlotTransform
Applies slot-level operations to packed (space-to-depth) latents by permuting/rotating the P x P slots inside the channel dimension. - Menu category: latent/channel - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Packed latent to transform per slot. | | operation | enum | shuffle | shuffle/rotatecw/rotateccw/fliph/flipv | Slot operation to apply. | | patchsize | INT | 2 | 1..8 | Patch size P for P x P slot layout. | | basechannels | INT | 0 | 0..4096 | Base channels before packing (0 = infer). | | seed | INT | 0 | 0..2^64-1 | Seed for slot shuffles. | | mix | FLOAT | 1.0 | 0.0..1.0 | See Common Parameters. | | matchstats | BOOLEAN | false | – | Match per-channel mean/std after edit. | | mask | MASK | – | – | Optional mask to limit the transform. | | Mode | What it does | Likely impact | |------|--------------|---------------| | shuffle | Randomly permutes the P x P slot ordering within each base channel. | Digital macro‑blocking, moiré‑like micro‑shifts, checkerboard artifacts. | | rotatecw | Rotates the slot grid clockwise (90 degrees). | Directional slot shifts; subtle “micro‑rotation” with structured aliasing. | | rotateccw | Rotates the slot grid counterclockwise (90 degrees). | Same as rotatecw but in opposite direction; can pair for oscillation. | | fliph | Flips slots left‑right. | Mirrored slot patterns; can cause chromatic‑aberration‑like shifts. | | flipv | Flips slots top‑bottom. | Vertical slot mirroring; structured digital shimmer. |
Latent/Channel · 1 Ausgabe
- LatentPerlinFractalNoise
LatentPerlinFractalNoise
Adds smooth fractal Perlin noise to a latent for structured variation. - Menu category: Latent/Noise - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent that will be perturbed with fractal Perlin noise. | | seed | INT | 0 | 0..2^64-1 | Seed controlling the procedural noise pattern. | | frequency | FLOAT | 2.0 | 0.01..64.0 | Base lattice frequency (higher = finer detail). | | octaves | INT | 4 | 1..12 | Number of noise layers to accumulate. | | persistence | FLOAT | 0.5 | 0.0..1.0 | Amplitude multiplier between octaves. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Frequency multiplier between octaves. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized noise relative to the latent's standard deviation. | | channelmode | enum | shared | shared/perchannel | See Common Parameters. | - For 5D latents (B,C,T,H,W), this node generates 3D Perlin noise across (T,H,W) (no temporalmode parameter).
Latent/Noise · 2 Eingaben · 7 Parameter · 1 Ausgabe
- LatentReactionDiffusion
LatentReactionDiffusion
Runs a Gray-Scott reaction-diffusion simulation and injects the resulting pattern into a latent. - Menu category: Latent/Noise - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent that will receive reaction-diffusion patterns. | | seed | INT | 0 | 0..2^64-1 | Seed for the initial chemical concentrations. | | iterations | INT | 200 | 1..2000 | Number of Gray-Scott simulation steps. | | feedrate | FLOAT | 0.036 | 0.0..0.1 | Feed rate (F) controlling how quickly chemical U is replenished. | | killrate | FLOAT | 0.065 | 0.0..0.1 | Kill rate (K) regulating removal of chemical V. | | diffusionu | FLOAT | 0.16 | 0.0..1.0 | Diffusion rate for chemical U. | | diffusionv | FLOAT | 0.08 | 0.0..1.0 | Diffusion rate for chemical V. | | timestep | FLOAT | 1.0 | 0.01..5.0 | Simulation time step used during integration. | | strength | FLOAT | 0.75 | 0.0..5.0 | Scales normalized pattern relative to the latent's standard deviation. | | channelmode | enum | shared | shared/perchannel | Reuse one simulation for all channels or rerun per channel. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Latent/Noise · 2 Eingaben · 10 Parameter · 1 Ausgabe
- LatentSimplexNoise
LatentSimplexNoise
Adds layered fractal simplex noise to a latent for organic perturbations. - Menu category: Latent/Noise - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to perturb with simplex noise. | | seed | INT | 0 | 0..2^64-1 | Seed controlling the simplex lattice offsets. | | frequency | FLOAT | 2.0 | 0.01..64.0 | Base lattice frequency for the simplex grid. | | octaves | INT | 4 | 1..12 | Number of simplex layers to accumulate. | | persistence | FLOAT | 0.5 | 0.0..1.0 | Amplitude multiplier applied between octaves. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Frequency multiplier applied between octaves. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized simplex noise relative to the latent's standard deviation. | | channelmode | enum | shared | shared/perchannel | See Common Parameters. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Latent/Noise · 2 Eingaben · 8 Parameter · 1 Ausgabe
- LatentSmokeSimulation
LatentSmokeSimulation
Simulates buoyancy-driven smoke in latent space and returns density and velocity previews. - Menu category: latent/perturb - Returns: LATENT, IMAGE, IMAGE See docs/fluid-simulation.md for smoke-mode parameters such as smokesourcemode, buoyancy, cooling, and output batching controls.
Latent/Perturb · 2 Eingaben · 21 Parameter · 3 Ausgaben
- LatentSwirlNoise
LatentSwirlNoise
Swirls latent pixels around randomized centers (vortex-like warp). This is a spatial deformation, not additive noise. - Menu category: Latent/Noise - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to deform with vortex-style warps. | | seed | INT | 0 | 0..2^64-1 | Seed for vortex placement and direction randomness. | | vortices | INT | 1 | 1..16 | Number of independent vortex centres to spawn per latent. | | channelmode | enum | global | global/perchannel | Shared swirl grid for all affected channels or unique grid per channel. | | channelfraction | FLOAT | 1.0 | 0.0..1.0 | Fraction of channels to swirl (subset chosen per sample). | | strength | FLOAT | 0.75 | 0.0..6.28 | Peak swirl rotation in radians near the vortex center. | | radius | FLOAT | 0.5 | 0.05..2.0 | Normalized radius controlling how far the vortex influence extends. | | centerspread | FLOAT | 0.25 | 0.0..1.0 | How far the vortex origin drifts from the latent center. | | directionbias | FLOAT | 0.0 | -1.0..1.0 | Bias toward counter-clockwise (1) or clockwise (-1) swirl. | | mix | FLOAT | 1.0 | 0.0..1.0 | See Common Parameters. |
Latent/Noise · 2 Eingaben · 9 Parameter · 1 Ausgabe
- LatentToBatch
LatentToBatch
Replaces or inserts a latent or latent batch in an existing LATENT batch at a selected batch index. - Menu category: latent/batch - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latentbatch | LATENT | - | - | Existing latent batch to modify. | | latent | LATENT | - | - | Latent or latent batch to place. Sample shape after the batch dimension, dtype, and device must match latentbatch. | | index | INT | 0 | 0..4096 | Target batch index. In replace mode it must point at an existing item; in insert mode it can also equal the current batch size to append. | | mode | enum | replace | replace/insert | replace overwrites existing item(s). insert shifts existing items right. | - noisemask is spliced when both inputs provide it; singleton masks are expanded to the matching sample count. - Existing batchindex metadata is preserved and spliced when present.
Latent/Batch · 2 Eingaben · 2 Parameter · 1 Ausgabe
- LatentToImage
LatentToImage
Kompakte Beschreibung für diesen Knoten nicht verfügbar.
Latent/Debug · 1 Eingabe · 2 Parameter · 1 Ausgabe
- LatentWorleyNoise
LatentWorleyNoise
Generates cellular (Worley) noise for cracked / biological / bubble-like textures. - Menu category: Latent/Noise - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Latent to perturb with Worley (cellular) noise. | | seed | INT | 0 | 0..2^64-1 | Seed for the feature point distribution. | | featurepoints | INT | 16 | 1..4096 | Base number of feature points scattered across the plane. | | octaves | INT | 3 | 1..8 | Number of cellular layers to accumulate. | | persistence | FLOAT | 0.65 | 0.0..1.0 | Amplitude multiplier between Worley octaves. | | lacunarity | FLOAT | 2.0 | 1.0..6.0 | Multiplier for the feature point count between octaves. | | distancemetric | enum | euclidean | euclidean/manhattan/chebyshev | Distance metric used when measuring feature proximity. | | jitter | FLOAT | 0.35 | 0.0..1.0 | How far feature points can drift inside each cell. | | strength | FLOAT | 0.5 | 0.0..5.0 | Scales normalized Worley noise relative to the latent's standard deviation. | | channelmode | enum | shared | shared/perchannel | See Common Parameters. | | temporalmode | enum | locked | locked/animated | See Common Parameters. |
Latent/Noise · 2 Eingaben · 10 Parameter · 1 Ausgabe
- ListSlice
ListSlice
Extracts a contiguous slice from a list-valued input without changing the item type. - Menu category: utils/list - Returns: list of the same item type | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | listitems | list | – | – | List-valued input to slice. Non-list inputs are copied through unchanged. | | startindex | INT | 1 | 1..2147483647 | 1-based starting item. Values past the list end clamp to the last item. | | length | INT | 1 | 1..2147483647 | Maximum number of items to include. Values past the list end truncate to available items. | | Output | Type | Description | |------|------|-------------| | listitems | list | The selected contiguous slice, or the original value when the input is not a list. | - startindex and length must both be 1 or greater. - Empty list inputs return an empty list.
Utils/List · 1 Eingabe · 2 Parameter · 1 Ausgabe
- MergeMustacheVariableDefs
MergeMustacheVariableDefs
Merges two MUSTACHEVARIABLES mappings into one. Use this when two branches define candidate pools and you want one combined variable-definition dictionary before Mustache Variable Sampler. - Menu category: text/template - Returns: MUSTACHEVARIABLES | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | variabledefs1 | MUSTACHEVARIABLES | – | – | First variable-definition mapping. | | variabledefs2 | MUSTACHEVARIABLES | – | – | Second variable-definition mapping. | | conflictmode | enum | keepfirst | keepfirst/keepsecond/mergevalues | Conflict policy for duplicate variable names. | | Mode | Effect | |------|------| | keepfirst | Preserve the value list from variabledefs1 and ignore the conflicting definition from variabledefs2. | | keepsecond | Overwrite the conflicting definition from variabledefs1 with the one from variabledefs2. | | mergevalues | Append the candidate values from variabledefs2 after the values from variabledefs1. | - This node merges variable definitions, not concrete sampled entries. If you already have MUSTACHEVARIABLELIST outputs, use Merge Mustache Variable Sets instead. - mergevalues follows the same duplicate-key semantics as multi-input Mustache Variables: it appends plain value lists in order, but conflicting weighted definitions still raise an error.
Text/Template · 1 Ausgabe
- MergeMustacheVariableSets
MergeMustacheVariableSets
Merges two MUSTACHEVARIABLELIST inputs entry-by-entry so each output entry contains the union of keys from both sides. This is the node to use when one branch produces fields like cameralens and another produces fields like personmasc, and your downstream template needs both in the same rendered entry. - Menu category: text/template - Returns: MUSTACHEVARIABLELIST | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | variablesets1 | MUSTACHEVARIABLELIST | – | – | First variable-setting list. | | variablesets2 | MUSTACHEVARIABLELIST | – | – | Second variable-setting list. Keys from this side overwrite duplicate keys from variablesets1. | | Output | Type | Description | |------|------|-------------| | variablesets | MUSTACHEVARIABLELIST | Entry-by-entry merged variable-setting list. | - If both inputs have the same length, entries are merged by index. - If one side has exactly one entry, that singleton entry is broadcast across the other side. - If both sides have different lengths greater than one, the node raises an error. - Use this instead of Concatenate Lists when downstream templates need one entry that contains keys from both branches.
Text/Template · 2 Eingaben · 0 Parameter · 1 Ausgabe
- ModelsList
ModelsList
Combines multiple MODEL inputs into a list for downstream list-mapped nodes such as KSampler.
Model/Batch · 2 Eingaben · 0 Parameter · 1 Ausgabe
- MustacheTemplate
MustacheTemplate
Renders a template string containing mustache placeholders like {{haircolor}} against a MUSTACHEVARIABLELIST and outputs one string per concrete variable-setting entry. Because the output is a ComfyUI STRING list, downstream nodes such as CLIP Text Encode will execute once per rendered prompt. - Menu category: text/template - Returns: STRING list | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | variablesets | MUSTACHEVARIABLELIST | – | – | Concrete variable settings generated by Mustache Variable Sampler. One output prompt is rendered for each list entry. | | template | STRING | "The man has {{haircolor}} hair and {{leglength}} legs." | multiline | Template text to render. Every referenced placeholder must exist in each MUSTACHEVARIABLELIST entry. Repeated placeholders reuse the same selected value within a given entry. | The node itself no longer computes permutations. Instead, it receives an already-expanded MUSTACHEVARIABLELIST from Mustache Variable Sampler and renders the template once per entry. For the example above and the template: text The man has {{haircolor}} hair and {{leglength}} legs. the node emits these six strings, in order: text The man has brown hair and short legs. The man has brown hair and long legs. The man has brown hair and weird legs. The man has blonde hair and short legs. The man has blonde hair and long legs. The man has blonde hair and weird legs. - Variables present in a MUSTACHEVARIABLELIST entry but not referenced in the template are ignored. - Templates with no placeholders are repeated…
Text/Template · 1 Eingabe · 1 Parameter · 1 Ausgabe
- MustacheVariable
MustacheVariable
Builds a one-key MUSTACHEVARIABLES mapping without writing YAML manually. Use it when you already have a variable name and either one string or a list-valued STRING input that should become that variable's candidate values. - Menu category: text/template - Returns: MUSTACHEVARIABLES | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | variablename | STRING | animal | – | Variable name to define. The name is trimmed, must be non-empty, and cannot use reserved/internal names. | | variablevalues | STRING or STRING list | fox | multiline | Scalar values are wrapped into a one-item candidate list. If a list-valued STRING input is connected, every item becomes one candidate value under variablename, in order. | - variablename = animal, variablevalues = fox returns {"animal": ["fox"]} - variablename = animal, variablevalues = ["fox", "wolf", "cat"] returns {"animal": ["fox", "wolf", "cat"]} - This node returns MUSTACHEVARIABLES, not MUSTACHEVARIABLELIST. Feed it into Mustache Variable Sampler when you want concrete per-entry settings. - It is useful as a bridge from list-valued text outputs into the mustache-variable pipeline without going through YAML.
Text/Template · 0 Eingaben · 2 Parameter · 1 Ausgabe
- MustacheVariables
MustacheVariables
Parses a YAML mapping of variable names to candidate values and packages it as a MUSTACHEVARIABLES dictionary for downstream templating nodes. For a workflow-oriented walkthrough with recursive-template patterns and operator usage, see docs/mustachetemplateguide.yaml. For a larger tracked YAML sample that follows that guide's layered approach and demonstrates weighted optional clauses, clause reuse, deeper nested composition, and semantically tuned sampled outputs, see sample.yaml. - Menu category: text/template - Returns: MUSTACHEVARIABLES | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | yamltext | STRING or STRING list | sample YAML | multiline | YAML mapping of variable names to values, or a YAML list of mappings. Lists of scalar values are the normal form for each variable; scalar values are accepted as shorthand and are wrapped into a one-item list. Each final value is coerced to text before rendering. When a list of YAML strings is provided, all mappings are merged together. | | variablesets | MUSTACHEVARIABLELIST | – | optional | Optional second-stage templating input. In the normal case you define variables directly in yamltext and reference them there. Connect this only when you want upstream concrete variable settings to render the YAML before it is parsed. | yaml haircolor: - brown - blonde leglength: - short - long - weird Variables are normally defined and referenced inside the same YAML block: yaml color: - brown - blue - black hairarrangement: - ponytail - bun hairstyle: - {{color:static}} hair in a {{haira…
Text/Template · 1 Eingabe · 1 Parameter · 1 Ausgabe
- MustacheVariableSampler
MustacheVariableSampler
Expands a MUSTACHEVARIABLES mapping into a MUSTACHEVARIABLELIST, where each entry is a concrete variable assignment like {"haircolor": "brown", "leglength": "short"}. - Menu category: text/template - Returns: MUSTACHEVARIABLELIST | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | variabledefs | MUSTACHEVARIABLES | – | – | Variable-definition mapping to expand into concrete per-prompt settings. | | samplingmode | enum | sequential | sequential/random | sequential walks the Cartesian product in stable order. random randomizes key order, value order, and sampled permutation order before emitting concrete settings. | | seed | INT | 0 | 0..2^64-1 | 64-bit seed used when samplingmode = random. The same seed yields the same concrete variable-setting order. | | limit | INT | -1 | -1..2147483647 | Maximum number of concrete variable settings to emit. -1 means no limit. | - This node now owns permutation generation, so large Cartesian products can be capped before Mustache Template runs. - Locally derived YAML values from Mustache Variables also expand here, not during YAML parsing, so sampler limit is now the place where recursive/local variable explosions are controlled. - Lazy placeholder settings like {{color:randomize}} are also resolved here. They use the node's seeded RNG, so repeated runs with the same seed stay deterministic. - Lazy variables are evaluated in dependency order rather than raw YAML insertion order, so YAML variable order no longer affects valid reference resolution. - In random mode, the sampler does not merel…
Text/Template · 1 Ausgabe
- MustacheVariableSet
MustacheVariableSet
Builds a single-entry MUSTACHEVARIABLELIST from one concrete key/value pair. Use it when you want to feed a simple value directly into Mustache Template without creating MUSTACHEVARIABLES definitions and sampling them first. - Menu category: text/template - Returns: MUSTACHEVARIABLELIST | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | variablename | STRING | animal | – | Variable name to define in the returned concrete variable set. The name is trimmed, must be non-empty, and cannot use reserved/internal names. | | variablevalue | STRING | fox | multiline | Concrete value to store under variablename. | | Output | Type | Description | |------|------|-------------| | variablesets | MUSTACHEVARIABLELIST | A one-item list containing one concrete variable-setting dictionary. | - variablename = animal, variablevalue = fox returns [{"animal": "fox"}] - This node returns MUSTACHEVARIABLELIST, not MUSTACHEVARIABLES, so it can connect directly to Mustache Template. - Use Mustache Variable instead when you want to define candidate values that should still pass through Mustache Variable Sampler.
Text/Template · 0 Eingaben · 2 Parameter · 1 Ausgabe
- NextSeeds
NextSeeds
Expands one 64-bit seed into four deterministic 64-bit seed outputs using a MurmurHash-style 64-bit mixer over offset streams. Each output can emit one seed or a list of seeds from its own deterministic stream. - Menu category: utils/seed - Returns: INT, INT, INT, INT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | seed | INT | 0 | 0..2^64-1 | Base ComfyUI seed. The node first treats it as an unsigned 64-bit integer, then derives four independent-looking child seeds from it. Use this when you want one master seed to drive four downstream nodes without manually picking related seed values. | | count | INT | 1 | 1..4096 | Number of seeds to emit per output. 1 preserves the original four-way fan-out. Larger values emit list-valued INT batches on each output. | | Output | Type | Description | |------|------|-------------| | seed1 | INT | First derived 64-bit seed stream. Emits one seed when count = 1, otherwise a list of count seeds. | | seed2 | INT | Second derived 64-bit seed stream. Emits one seed when count = 1, otherwise a list of count seeds. | | seed3 | INT | Third derived 64-bit seed stream. Emits one seed when count = 1, otherwise a list of count seeds. | | seed4 | INT | Fourth derived 64-bit seed stream. Emits one seed when count = 1, otherwise a list of count seeds. | Starting from the input seed, the node creates 64-bit stream values by adding a large fixed increment each time (0x9E3779B97F4A7C15, a golden-ratio-derived step commonly used for sequence spacing in hash/PRNG code). Each stream value is then passed t…
Utils/Seed · 0 Eingaben · 2 Parameter · 4 Ausgaben
- PatchifyFlux2Latent
PatchifyFlux2Latent
Re-patchifies an unpatchified Flux.2 latent back to the standard 2x2 patch format. - Menu category: Latent/Flux - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Unpatchified Flux.2 latent to return to 128-channel patchified form. | - Use this after latent noise nodes to restore the format expected by Flux.2 models and downstream nodes.
Latent/Flux · 1 Eingabe · 0 Parameter · 1 Ausgabe
- ReorderList
ReorderList
Reorders a list-valued input without changing the item type. This is a generic utility node that can reverse any list or apply a seeded shuffle. - Menu category: utils/list - Returns: list of the same item type | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | listitems | list | – | – | List-valued input to reorder. | | mode | enum | shuffle | shuffle/reverse | shuffle applies a seeded random permutation. reverse flips the list order. | | seed | INT | 0 | 0..2^64-1 | 64-bit seed used when mode = shuffle. The same seed yields the same permutation. | | Output | Type | Description | |------|------|-------------| | listitems | list | Reordered list of the same item type as the input. | - This node is list-aware and intended for list-valued sockets such as STRING lists coming out of Mustache Template. - reverse ignores seed.
Utils/List · 1 Ausgabe
- RotorQuantAttentionModelPatch
RotorQuantAttentionModelPatch
Kompakte Beschreibung für diesen Knoten nicht verfügbar.
Model/Patch · 1 Ausgabe
- SplitLatentPhaseMagnitude
SplitLatentPhaseMagnitude
Converts a spatial latent into two frequency-domain latents: magnitude and phase.
Latent/Frequency · 1 Eingabe · 0 Parameter · 2 Ausgaben
- TurboQuantAttentionModelPatch
TurboQuantAttentionModelPatch
Kompakte Beschreibung für diesen Knoten nicht verfügbar.
Model/Patch · 1 Ausgabe
- UnpatchifyFlux2Latent
UnpatchifyFlux2Latent
Converts patchified Flux.2 latents into unpatchified latents with double spatial resolution. - Menu category: Latent/Flux - Returns: LATENT | Field | Type | Default | Range/Options | Notes | |------|------|---------|--------------|------| | latent | LATENT | – | – | Flux.2 latent to expand from 128-channel 2x2 patchified form. | - Use this before latent noise nodes so their spatial operations happen on true neighboring pixels rather than 2x2 patch blocks.
Latent/Flux · 1 Eingabe · 0 Parameter · 1 Ausgabe
Quellen
1 Quelle
Quellen
1 QuelleQuellenauszüge
1 Auszug
Quellenauszüge
1 AuszugSource context: Repo ttulttul/Skoogeer-Noise
