Sound and voice¶

A video can have sound: a sound with an animation, music under the whole video, or a voice that explains. manimgx puts each sound at its moment in the video, and mixes them into one sound track.

Play a sound with an animation¶

Sound reads a sound from a file: WAV, MP3, AAC, FLAC and the other usual formats. Give it to self.play with an animation, and it starts with the animation:

self.play(m.FadeIn(dot), m.Sound("click.wav"))

To start a sound at any moment, without an animation, use self.add_sound: the sound plays while the scene continues.

self.add_sound(m.Sound("theme.mp3"))

A sound plays to its end. In self.play, the play lasts until the sound ends; self.add_sound doesn't wait. To stop a sound before its end, keep what add_sound gives back: music = self.add_sound(...), then music.stop(fade=2).

Change a sound¶

A sound's methods make a changed sound: here, music 18 decibels quieter, which fades in for 2 seconds and plays until the video ends. A sound that loops never ends, so start it with add_sound: self.play refuses it.

music = self.add_sound(m.Sound("theme.mp3").gain(-18).fade_in(2).loop())

The Sound reference shows every change.

Say a line¶

say speaks a line with the scene's voice, and plays animations while it speaks. Put the words of each animation in brackets, and give the animations in the same order. The voice doesn't say the brackets:

import manimgx as m


class Derivative(m.Scene):
    def construct(self) -> None:
        square = m.MathTex("x^2", font_size=96).shift(2 * m.LEFT)
        derivative = m.MathTex("2x", font_size=96).shift(2 * m.RIGHT)
        self.add(square)
        self.say(
            "The derivative of [x squared] is [two x].",
            m.Indicate(square),
            m.Write(derivative),
        )

Each animation starts with the first of its words. The scene waits until the line ends, so lines don't overlap. If you change the words, the animations follow the new words.

Voices¶

The default voice is ElevenLabs' Eleven v3, on fal.ai. It needs your fal.ai key in the FAL_KEY environment variable. To choose a different voice, set voice in the scene's class, next to construct:

voice = m.voices.Fal(
    "fal-ai/minimax/speech-2.8-hd",
    voice_setting={"voice_id": "Calm_Woman", "emotion": "happy"},
)

Fal speaks with five models, and your editor shows the settings of each one. Any other text-to-speech service can be a voice too: a voice is a function from text to a Speech. For a dialogue, give say a voice for one line: self.say("Hello!", voice=other).

manimgx keeps what the voice says in a folder named voice/, next to your scene's file. So the scene renders again without the service, even offline, and only a changed line goes to the service again. To use your own recording for a line, put it in place of that line's audio, with the same name.

Captions¶

What the voice says becomes the video's captions. manimgx render writes them next to the video, as subtitles (.srt) that video players and video sites read. add_subcaption adds captions of your own.

Next¶

Rendering and sharing shows the ways to make your video, check it, and share it.