Skip to content

noise

Classes:

  • GrainFactoryBicubic

    Bicubic scaler originally implemented in GrainFactory with a sharp parameter.

  • Grainer

    Enum representing different grain/noise generation algorithms.

  • ScalerTwoPasses

    Scaler class that applies scaling in two passes.

Attributes:

LanczosTwoPasses module-attribute

LanczosTwoPasses = ScalerTwoPasses[Lanczos]

Lanczos resizer that applies scaling in two passes.

EdgeLimits

EdgeLimits = tuple[
    float | Sequence[float] | bool, float | Sequence[float] | bool
]

Tuple representing lower and upper edge limits for each plane.

Format: (low, high)

Each element can be:

  • A float: the same limit is applied to all planes.
  • A sequence of floats: individual limits for each plane.
  • True: use the default legal range per plane.
  • False: no limits are applied.

GrainerLike

GrainerLike = Grainer | GrainerPartial

Grainer-like type, which can be a single grainer or a partial grainer.

AbstractGrainer

Abstract grainer base class.

Methods:

  • __call__

    To be implemented in subclasses.

__call__

__call__(clip: VideoNode, /, **kwargs: Any) -> VideoNode | GrainerPartial

To be implemented in subclasses.

Source code in vsdeband/noise.py
131
132
133
def __call__(self, clip: vs.VideoNode, /, **kwargs: Any) -> vs.VideoNode | GrainerPartial:
    """To be implemented in subclasses."""
    raise NotImplementedError

GrainFactoryBicubic

GrainFactoryBicubic(sharp: float = 50, **kwargs: Any)

Bases: BicubicAuto

Bicubic scaler originally implemented in GrainFactory with a sharp parameter.

Initialize the scaler with optional arguments.

Parameters:

  • sharp

    (float, default: 50 ) –

    Sharpness of the scaler. Defaults to 50 which corresponds to Catrom scaling.

  • **kwargs

    (Any, default: {} ) –

    Keyword arguments that configure the internal scaling behavior.

Source code in vsdeband/noise.py
115
116
117
118
119
120
121
122
123
def __init__(self, sharp: float = 50, **kwargs: Any) -> None:
    """
    Initialize the scaler with optional arguments.

    Args:
        sharp: Sharpness of the scaler. Defaults to 50 which corresponds to Catrom scaling.
        **kwargs: Keyword arguments that configure the internal scaling behavior.
    """
    super().__init__(sharp / -50 + 1, None, **kwargs)

Grainer

Bases: AbstractGrainer, CustomEnum

Enum representing different grain/noise generation algorithms.

Methods:

  • __call__

    Apply grain to a clip using the selected graining method.

  • norm_brightness

    Normalize the brightness of the grained clip to match the original clip's average luminance.

Attributes:

FBM_SIMPLEX class-attribute instance-attribute

FBM_SIMPLEX = 3

Fractional Brownian Motion based on Simplex noise. Built-in vs-noise plugin.

GAUSS class-attribute instance-attribute

GAUSS = 0

Gaussian noise. Built-in vs-noise plugin.

PERLIN class-attribute instance-attribute

PERLIN = 1

Perlin noise. Built-in vs-noise plugin.

PLACEBO class-attribute instance-attribute

PLACEBO = auto()

Grain effect provided by the libplacebo rendering library.

POISSON class-attribute instance-attribute

POISSON = 4

Poisson-distributed noise. Built-in vs-noise plugin.

SIMPLEX class-attribute instance-attribute

SIMPLEX = 2

Simplex noise. Built-in vs-noise plugin.

__call__

__call__(
    clip: VideoNode,
    /,
    strength: float | tuple[float, float] = ...,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any,
) -> VideoNode
__call__(
    *,
    strength: float | tuple[float, float] = ...,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any
) -> GrainerPartial
__call__(
    clip: VideoNode,
    /,
    strength: float | tuple[float, float] = ...,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    *,
    size: int | tuple[float | None, float | None] | None = (2.0, 2.0),
    **kwargs: Any,
) -> VideoNode
__call__(
    *,
    strength: float | tuple[float, float] = ...,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    size: int | tuple[float | None, float | None] | None = (2.0, 2.0),
    **kwargs: Any
) -> GrainerPartial
__call__(
    clip: VideoNode,
    /,
    strength: float | Sequence[float] = ...,
    *,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any,
) -> VideoNode
__call__(
    *,
    strength: float | Sequence[float] = ...,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any
) -> GrainerPartial
__call__(
    clip: VideoNode,
    /,
    strength: float | tuple[float, float] = ...,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any,
) -> VideoNode
__call__(
    *,
    strength: float | tuple[float, float] = ...,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any
) -> GrainerPartial
__call__(
    clip: VideoNode | MissingT = MISSING,
    /,
    strength: float | Sequence[float] = 0,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: (
        Callable[[VideoNode], VideoNode]
        | Iterable[Callable[[VideoNode], VideoNode]]
        | None
    ) = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any,
) -> VideoNode | GrainerPartial

Apply grain to a clip using the selected graining method.

If no clip is passed, a partially applied grainer with the provided arguments is returned instead.

Example usage
# For PERLIN, SIMPLEX, and FBM_SIMPLEX, it is recommended to use `size` instead of `scale`,
# as `size` allows for direct internal customization of each grain type.
grained = Grainer.PERLIN(clip, (1.65, 0.65), temporal=(0.25, 2), luma_scaling=4, size=3.0, seed=333)

Parameters:

  • clip

    (VideoNode | MissingT, default: MISSING ) –

    The input clip to apply grain to. If omitted, returns a partially applied grainer.

  • strength

    (float | Sequence[float], default: 0 ) –

    Grain strength. A single float applies uniform strength to all planes. A sequence allows per-plane control.

  • static

    (bool, default: False ) –

    If True, the grain pattern is static (unchanging across frames).

  • scale

    (float | tuple[float, float], default: 1.0 ) –

    Scaling divisor for the grain layer. Can be a float (uniform scaling) or a tuple (width, height scaling).

  • scaler

    (ScalerLike, default: LanczosTwoPasses ) –

    Scaler used to resize the grain layer when scale is not 1.0.

  • temporal

    (float | tuple[float, int], default: (0.0, 0) ) –

    Temporal grain smoothing parameters. Either a float (weight) or a tuple of (weight, radius).

  • post_process

    (Callable[[VideoNode], VideoNode] | Iterable[Callable[[VideoNode], VideoNode]] | None, default: None ) –

    One or more functions applied after grain generation (and temporal smoothing, if used).

  • protect_edges

    (bool | EdgeLimits, default: True ) –

    Protects edge regions of each plane from graining.

    • True: Use legal range based on clip format.
    • False: Disable edge protection.
    • Tuple: Specify custom edge limits per plane (see EdgeLimits).
  • protect_neutral_chroma

    (bool | None, default: None ) –

    Whether to disable graining on neutral chroma.

  • luma_scaling

    (float | None, default: None ) –

    Sensitivity of the luma-adaptive graining mask. Higher values reduce grain in brighter areas; negative values invert behavior.

  • planes

    (Planes, default: None ) –

    Which planes to process. Default to all.

  • **kwargs

    (Any, default: {} ) –

    Additional arguments to pass to the graining function or additional advanced options:

    • temporal_avg_func: Temporal average function to use instead of the default standard mean.
    • protect_edges_blend: Blend range (float) to soften edge protection thresholds.
    • protect_neutral_chroma_blend: Blend range (float) for neutral chroma protection.
    • neutral_out: (Boolean) Output the neutral layer instead of the merged clip.

Returns:

Source code in vsdeband/noise.py
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
def __call__(
    self,
    clip: vs.VideoNode | MissingT = MISSING,
    /,
    strength: float | Sequence[float] = 0,
    static: bool = False,
    scale: float | tuple[float, float] = 1.0,
    scaler: ScalerLike = LanczosTwoPasses,
    temporal: float | tuple[float, int] = (0.0, 0),
    post_process: Callable[[vs.VideoNode], vs.VideoNode]
    | Iterable[Callable[[vs.VideoNode], vs.VideoNode]]
    | None = None,
    protect_edges: bool | EdgeLimits = True,
    protect_neutral_chroma: bool | None = None,
    luma_scaling: float | None = None,
    planes: Planes = None,
    **kwargs: Any,
) -> vs.VideoNode | GrainerPartial:
    """
    Apply grain to a clip using the selected graining method.

    If no clip is passed, a partially applied grainer with the provided arguments is returned instead.

    Example usage:
        ```py
        # For PERLIN, SIMPLEX, and FBM_SIMPLEX, it is recommended to use `size` instead of `scale`,
        # as `size` allows for direct internal customization of each grain type.
        grained = Grainer.PERLIN(clip, (1.65, 0.65), temporal=(0.25, 2), luma_scaling=4, size=3.0, seed=333)
        ```

    Args:
        clip:
            The input clip to apply grain to. If omitted, returns a partially applied grainer.
        strength:
            Grain strength.
            A single float applies uniform strength to all planes.
            A sequence allows per-plane control.

        static:
            If True, the grain pattern is static (unchanging across frames).
        scale:
            Scaling divisor for the grain layer.
            Can be a float (uniform scaling) or a tuple (width, height scaling).
        scaler:
            Scaler used to resize the grain layer when `scale` is not 1.0.
        temporal:
            Temporal grain smoothing parameters. Either a float (weight) or a tuple of (weight, radius).
        post_process:
            One or more functions applied after grain generation (and temporal smoothing, if used).
        protect_edges:
            Protects edge regions of each plane from graining.

               - True: Use legal range based on clip format.
               - False: Disable edge protection.
               - Tuple: Specify custom edge limits per plane (see [EdgeLimits][vsdeband.noise.EdgeLimits]).

        protect_neutral_chroma:
            Whether to disable graining on neutral chroma.
        luma_scaling:
            Sensitivity of the luma-adaptive graining mask.
            Higher values reduce grain in brighter areas; negative values invert behavior.
        planes:
            Which planes to process. Default to all.
        **kwargs:
            Additional arguments to pass to the graining function or additional advanced options:

               - ``temporal_avg_func``: Temporal average function to use instead of the default standard mean.
               - ``protect_edges_blend``: Blend range (float) to soften edge protection thresholds.
               - ``protect_neutral_chroma_blend``: Blend range (float) for neutral chroma protection.
               - ``neutral_out``: (Boolean) Output the neutral layer instead of the merged clip.

    Returns:
        Grained video clip, or a [GrainerPartial][vsdeband.noise.GrainerPartial] if `clip` is not provided.
    """
    kwargs.update(
        strength=strength,
        scale=scale,
        scaler=scaler,
        temporal=temporal,
        protect_edges=protect_edges,
        post_process=post_process,
        protect_neutral_chroma=protect_neutral_chroma,
        luma_scaling=luma_scaling,
        planes=planes,
    )

    if clip is MISSING:
        return GrainerPartial(self, **kwargs)

    if self == Grainer.PLACEBO:
        assert static is False, "PlaceboGrain does not support static noise!"

        return _apply_grainer(
            clip,
            lambda clip, strength, planes, **kwds: placebo_deband(
                clip, 8, 0.0, strength, planes, iterations=1, **kwds
            ),
            **kwargs,
            func=self.name,
        )

    if not isinstance(size := kwargs.pop("size", (None, None)), tuple):
        size = (size, size)

    kwargs.update(xsize=size[0], ysize=size[1])

    def _noise_function(
        clip: vs.VideoNode, strength: float | Sequence[float], planes: Planes, **kwds: Any
    ) -> vs.VideoNode:
        strength = normalize_param_planes(clip, strength, planes, 0)

        if len(set(strength[1:])) != 1:
            raise CustomValueError("Inconsistent grain values on chroma planes.", self.name, strength[1:])

        return core.noise.Add(clip, strength[0], strength[1], type=self.value, constant=static, **kwds)

    return _apply_grainer(clip, _noise_function, **kwargs, func=self.name)

norm_brightness staticmethod

norm_brightness() -> Callable[[VideoNode], VideoNode]

Normalize the brightness of the grained clip to match the original clip's average luminance.

Designed for use in the post_process parameter of Grainer().

Returns:

  • Callable[[VideoNode], VideoNode]

    A function that takes a grained clip and returns a brightness-normalized version.

Source code in vsdeband/noise.py
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
@staticmethod
def norm_brightness() -> Callable[[vs.VideoNode], vs.VideoNode]:
    """
    Normalize the brightness of the grained clip to match the original clip's average luminance.

    Designed for use in the `post_process` parameter of [Grainer()][vsdeband.Grainer.__call__].

    Returns:
        A function that takes a grained clip and returns a brightness-normalized version.
    """

    def _funtion(grained: vs.VideoNode) -> vs.VideoNode:
        for i in range(grained.format.num_planes):
            grained = core.std.PlaneStats(grained, plane=i, prop=f"PS{i}")

        if grained.format.sample_type is vs.FLOAT:
            expr = "x x.PS{plane_idx}Average -"
        else:
            expr = "x neutral range_size / x.PS{plane_idx}Average - range_size * +"

        return norm_expr(grained, expr, func=Grainer.norm_brightness)

    return _funtion

GrainerPartial

GrainerPartial(grainer: Grainer, **kwargs: Any)

Bases: AbstractGrainer

A partially-applied grainer wrapper.

Stores a grainer function, allowing it to be reused with different clips.

Parameters:

  • grainer

    (Grainer) –

    Grainer enumeration.

  • **kwargs

    (Any, default: {} ) –

    Arguments for the specified grainer.

Methods:

  • __call__

    Apply the grainer to the given clip with optional argument overrides.

Source code in vsdeband/noise.py
631
632
633
634
635
636
637
638
639
640
def __init__(self, grainer: Grainer, **kwargs: Any) -> None:
    """
    Stores a grainer function, allowing it to be reused with different clips.

    Args:
        grainer: [Grainer][vsdeband.noise.Grainer] enumeration.
        **kwargs: Arguments for the specified grainer.
    """
    self._grainer = grainer
    self._kwargs = kwargs

__call__

__call__(clip: VideoNode, /, **kwargs: Any) -> VideoNode

Apply the grainer to the given clip with optional argument overrides.

Parameters:

  • clip

    (VideoNode) –

    Clip to be processed.

  • **kwargs

    (Any, default: {} ) –

    Additional keyword arguments to override or extend the stored ones.

Returns:

  • VideoNode

    Processed clip.

Source code in vsdeband/noise.py
642
643
644
645
646
647
648
649
650
651
652
653
def __call__(self, clip: vs.VideoNode, /, **kwargs: Any) -> vs.VideoNode:
    """
    Apply the grainer to the given clip with optional argument overrides.

    Args:
        clip: Clip to be processed.
        **kwargs: Additional keyword arguments to override or extend the stored ones.

    Returns:
        Processed clip.
    """
    return self._grainer(clip, **self._kwargs | kwargs)

ScalerTwoPasses

Bases: ScalerSpecializer[_ScalerWithLanczosDefaultT]

Scaler class that applies scaling in two passes.

Methods:

scale

scale(
    clip: VideoNode,
    width: int | None = None,
    height: int | None = None,
    shift: tuple[TopShift, LeftShift] = (0, 0),
    **kwargs: Any
) -> VideoNode
Source code in vsdeband/noise.py
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
def scale(
    self,
    clip: vs.VideoNode,
    width: int | None = None,
    height: int | None = None,
    shift: tuple[TopShift, LeftShift] = (0, 0),
    **kwargs: Any,
) -> vs.VideoNode:
    assert check_variable_resolution(clip, self.__class__)

    if any(shift):
        raise CustomValueError("Shifting is unsupported.", self.__class__, shift)

    width, height = self._wh_norm(clip, width, height)

    if width / clip.width > 1.5 or height / clip.height > 1.5:
        # If the scale is too big, we need to scale it in two passes, else the window
        # will be too big and the grain will be dampened down too much
        mod = max(clip.format.subsampling_w, clip.format.subsampling_h) << 1
        clip = super().scale(
            clip, mod_x((width + clip.width) / 2, mod), mod_x((height + clip.height) / 2, mod), **kwargs
        )

    return super().scale(clip, width, height, (0, 0), **kwargs)