Extending Algan

Before diving into Algan’s internals to create a new class from scratch, check whether your goal can be achieved with existing extension hooks:

What you want

What to use

A custom mathematical animation

animated_function() (Custom Animations)

Continuous real-time behavior

Updaters (Updaters)

A new parametric 3-D surface

Surface with a custom coordinate function

Custom lighting / shading effects

Materials or fragment shader pipelines (Shaders and Materials)

Custom post-processing image filters

Post-processing passes (Backgrounds and Post-Processing)

External 2-D/3-D assets

ManimMob or Model3D

Building Custom Mob Classes

If you need a new shape, subclass an existing Mob base class:

  • 2-D Outlines & Shapes: Subclass BezierCircuitCubic. (See Circle or Polygon for reference implementations).

  • 3-D Curved Surfaces: Subclass Surface and provide a parametric coordinate function (u, v) -> (x, y, z).

  • Polyhedral Meshes: Subclass Polyhedron and supply vertex and face arrays.

Registering Custom Animatable Attributes

If your custom Mob class introduces new attributes that should animate smoothly over time on the timeline, register them during __init__:

class Ribbon(Surface):
    def __init__(self, twist=0.0, **kwargs):
        super().__init__(self._shape_func, **kwargs)
        # Register custom attribute with the timeline engine
        self.register_attrs_as_animatable(["twist"])
        self.twist = twist

Important

Always call self.register_attrs_as_animatable(["attr_name"]) before assigning the initial value in __init__.

Batching Large Numbers of Mobs

If your scene features hundreds or thousands of identical shapes (like a point cloud or a large particle grid), don’t create thousands of separate Mob instances. Instead, pack them into a single batched Mob:

Example: ExtendingPackedSpheres 

from algan import *
import torch

grid = torch.arange(12) * 0.5 - 2.75
centers = torch.stack(torch.meshgrid(grid, grid, indexing='ij'), -1)
centers = torch.cat((centers.reshape(-1, 2),
                     torch.zeros(centers.numel() // 2, 1)), -1)

# Builds 144 spheres in a single packed batch
spheres = Sphere.from_batches(centers, radius=0.15, color=BLUE).spawn()

spheres.move(UP * 0.5)   # Moves all 144 spheres at once
spheres[7].move(OUT)     # Moves an individual sphere view

Scene.save_video()

This stores all items in a single contiguous GPU buffer, keeping rendering memory and CPU timeline overhead minimal.

Contributing Extensions

If you create new shapes or animations that would benefit other users, we’d love to review your PR! Check out Contributing and Development for development setup and testing guidelines.

See Also