The animation engine for agents: blazingly fast math and 3D videos with Manim's Python API, on a Rust GPU core.
Python
Camera and 3D
The camera decides what the video shows. Its frame is a rectangle in the scene: the frame that Positions measures. Move the frame, and the view moves with it.
Pan and zoom
self.camera.frame is a mobject. Animate it as any other mobject:
import manimgx as m
class Zoom(m.Scene):
def construct(self) -> None:
circle = m.Circle(radius=3, color=m.BLUE)
tangent = m.Line(4 * m.LEFT, 4 * m.RIGHT, color=m.YELLOW)
tangent.shift(3 * m.UP)
self.add(circle, tangent)
frame = self.camera.frame
self.play(frame.animate.scale(0.25).move_to(3 * m.UP), run_time=2)
self.wait()
self.play(frame.animate.scale(4).move_to(m.ORIGIN), run_time=2)
- Move the frame to pan:
frame.animate.move_to(point). - Scale it to zoom: a frame half as large shows everything twice as large.
Three dimensions
To make a scene in 3D, make it from m.ThreeDScene. Its camera looks at the frame from
two angles:
phi, its angle from straight above: 0 looks straight down, as in a flat scene.theta, its angle around the vertical axis.
import manimgx as m
import numpy as np
class Hill(m.ThreeDScene):
def construct(self) -> None:
axes = m.ThreeDAxes(
x_range=[-3, 3],
y_range=[-3, 3],
z_range=[0, 2],
x_length=7,
y_length=7,
z_length=3,
)
def height(u: float, v: float) -> np.ndarray:
return axes.c2p(u, v, 2 * np.exp(-(u**2 + v**2) / 2))
hill = m.Surface(height, u_range=[-2.5, 2.5], v_range=[-2.5, 2.5])
title = m.Text("A hill", font_size=40)
title.to_corner(m.UL)
self.add_fixed_in_frame_mobjects(title)
self.add(axes, hill)
self.move_camera(phi=65 * m.DEGREES, theta=-45 * m.DEGREES, run_time=2)
self.begin_ambient_camera_rotation(rate=0.3)
self.wait(3)
self.set_camera_orientation(phi=..., theta=...)turns the camera at once, andself.move_camera(...)turns it in an animation.self.begin_ambient_camera_rotation(rate=0.3)turns the camera around the scene, untilself.stop_ambient_camera_rotation().self.add_fixed_in_frame_mobjects(title)keeps a mobject still on the screen while the camera moves: here, the title stays in its corner.self.remove_fixed_in_frame_mobjects(title)lets the camera move it again.
Every mobject has three coordinates, so flat mobjects work in 3D too. The [reference][manimgx.mobjects.three_d] shows the mobjects made for 3D, such as spheres and cubes.
Next
Sound and voice shows how to add sound and a voice to your video.