3b1b/manim · error · Exception

Matrix has bad dimensions

Error message

Matrix has bad dimensions

What it means

ApplyMatrix animates a linear transformation of a mobject's points. initialize_matrix accepts exactly a (2,2) matrix (promoted to 3x3 in the xy-plane) or a (3,3) matrix; anything else — 1x3, 4x4, non-square, wrong dtype shape — raises 'Matrix has bad dimensions'.

Source

Thrown at manimlib/animation/transform.py:288

        matrix: npt.ArrayLike,
        mobject: Mobject,
        **kwargs
    ):
        matrix = self.initialize_matrix(matrix)

        def func(p):
            return np.dot(p, matrix.T)

        super().__init__(func, mobject, **kwargs)

    def initialize_matrix(self, matrix: npt.ArrayLike) -> np.ndarray:
        matrix = np.array(matrix)
        if matrix.shape == (2, 2):
            new_matrix = np.identity(3)
            new_matrix[:2, :2] = matrix
            matrix = new_matrix
        elif matrix.shape != (3, 3):
            raise Exception("Matrix has bad dimensions")
        return matrix


class ApplyComplexFunction(ApplyMethod):
    def __init__(
        self,
        function: Callable[[complex], complex],
        mobject: Mobject,
        **kwargs
    ):
        self.function = function
        method = mobject.apply_complex_function
        super().__init__(method, function, **kwargs)

    def init_path_func(self) -> None:
        func1 = self.function(complex(1))
        self.path_arc = np.log(func1).imag
        super().init_path_func()

View on GitHub (pinned to dee01804d4)

Solutions

1. Reshape to exactly 2x2 or 3x3: np.array(mat).reshape(3, 3)
2. For homogeneous transforms, drop the last row and column before passing
3. For pure rotations, prefer ApplyMethod(mob.rotate, angle, axis=Z) or Rotation(angle) which handle the matrix for you

Example fix

# before
self.play(ApplyMatrix(rotation_matrix_4x4, mob))
# after
self.play(ApplyMatrix(rotation_matrix_4x4[:3, :3], mob))

import numpy as np
mat = np.asarray(mat)
assert mat.shape in ((2, 2), (3, 3)), f"matrix must be 2x2 or 3x3, got {mat.shape}"

import numpy as np
def is_valid_transform_matrix(mat) -> bool:
    return np.asarray(mat).shape in ((2, 2), (3, 3))

When it happens

Trigger: ApplyMatrix([[1, 0, 0], [0, 1, 0]], mob) (2x3); passing a rotation_about_z(...) result plus an extra row; passing a 4x4 homogeneous matrix from graphics code; passing a flat list of 9 numbers without reshaping.

Common situations: Mixing up homogeneous 4x4 matrices from other graphics libraries; building matrices by hand and getting row counts wrong; intending ApplyPointwiseFunction or apply_matrix on the mobject instead.

Related errors

• {self.__class__.__name__}.create_target not properly impleme
• MoveToTarget called on mobject without attribute 'target'
• Whoops, looks like you accidentally invoked the method you w
• Trying to restore without having saved
• Functions passed to ApplyFunction must return object of type

AI-assisted analysis of 3b1b/manim@dee01804d4 (2026-08-14). Data as JSON: /api/errors/3b0c7754fda6e39c. Report an issue: GitHub.