Audio And Sound Effects

Choose An Audio Channel

Audio is classified by purpose and timing. Pick the channel first; the renderer keeps its metadata separate in the editor and combines all active channels into the final MP4 mix.

Voiceovers

Use a voiceover scene class when spoken narration should own pacing or captions. Define stable keys withset_voiceovers, then play a key in a context. The tracker exposes the measuredduration, and the scene is padded when narration outlasts the animations in its block.

from manim import * class NarratedOrbitScene(VoiceoverThreeDScene):    def construct(self):        self.set_voiceovers({            "intro": {                "text": "Reusable narration on top of a Three D scene.",                "mp3": "/data/audio/intro.mp3",            },        })         sphere = Sphere(radius=1.2)        with self.play_voiceover_block("intro") as tracker:            self.play(Create(sphere), run_time=min(1.0, tracker.duration)) 

Choose VoiceoverScene, VoiceoverMovingCameraScene, orVoiceoverThreeDScene to match the visual scene type. Upload or generate narration first and bind the resulting workspace path. Do not use bundled sound effects for speech.

Background Audio

Use background audio for music or ambience that should cover a known scene span. It works on base and voiceover scene classes. The source is an uploaded workspace asset or workspace-confined audio path.

from manim import * class BackgroundAudioScene(Scene):    def construct(self):        with self.play_background_audio_block(            "/data/audio/theme.mp3",            volume=0.25,            fade_in=0.2,            fade_out=0.3,            loop=True,        ):            self.play(Create(Circle()), run_time=2)            self.wait(1) 

loop=True repeats the source to cover the block. Without looping, playback stops when the source ends. A leading slash means the root of the current user workspace—not the server filesystem. Targets outside that workspace and symlink escapes are rejected.

Sound Effects

Use self.play_sound for a trusted, package-owned sound, including both short effects and the longer documentary tracks below. It works on Scene, ThreeDScene, moving-camera scenes, and every voiceover variant. It does not accept uploaded files, paths, URLs, or raw audio data—only an exact catalog ID.

from manim import * class SoundEffectScene(Scene):    def construct(self):        button = RoundedRectangle(width=3.2, height=1.2)        label = Text("Launch").move_to(button)        self.add(button, label)         # Without duration, the with-body is the playback boundary. A longer        # sound is trimmed; a shorter sound ends naturally.        with self.play_sound("ui/click", volume=0.8) as cue:            self.play(button.animate.set_fill(BLUE, opacity=0.35), run_time=0.4)         with self.play_sound("movement/slide", fade_out=0.05):            self.play(VGroup(button, label).animate.shift(UP * 2), run_time=0.8)         # duration requests an audible span. Extending beyond the source        # requires loop=True.        with self.play_sound(            "ui/click",            duration=1.2,            loop=True,            fade_in=0.15,            fade_out=0.25,        ):            self.play(button.animate.set_stroke(YELLOW), run_time=1.2)         # The sound can continue over later scene work without padding this        # block when asynchronous timing is intentional.        with self.play_sound("ui/confirm", duration=0.5, pad_block=False):            self.wait(0.02) 

Without an explicit duration, the with-body is authoritative: a longer source is trimmed, a shorter source ends naturally, and loop=True repeats a shorter source through the body. This default never adds a sound-driven wait. An explicit duration sets the requested playback span and may outlast the body. If it also outlasts the source, only loop=True repeats the source; otherwise playback stops at the source's natural end with no fake silent tail. pad_block=True waits for explicit playback that actually outlasts the body, while pad_block=False lets it continue asynchronously.

Nature effects

from manim import * class NatureSoundScene(Scene):    def construct(self):        feather = TablerIcon("feather", color=YELLOW).scale(1.2)        drop = Dot(color=BLUE).shift(UP * 2)         with self.play_sound("nature/wind/gust", volume=0.7):            self.play(FadeIn(feather), run_time=0.5)         with self.play_sound("nature/water/drop"):            self.play(drop.animate.shift(DOWN * 4), run_time=0.7)         with self.play_sound("nature/weather/thunder", fade_in=0.05):            self.play(Flash(ORIGIN, color=WHITE), run_time=0.4) 

This nature example is also kept as the runnable renderer fixturepackages/manim_renderer/tests/fixtures/nature_sound_effect_scene.py so the documented API is exercised by real MP4 verification.

Unknown well-formed IDs produce a compile error with close-name suggestions. Malformed input—including an extension, whitespace, traversal segment, URL, markup, shell fragment, non-string value, or string subclass—is rejected before FFmpeg runs. Without an explicit duration, playback depends on the completed with-body, so fade fit is validated when that body exits.

Documentary Library

The documentary/* namespace is a built-in CC0 collection for serious science, space, technology, and reflective storytelling. Use it only through play_sound. The separate background-audio API continues to resolve user-owned workspace audio and does not expose these packaged files.

from manim import * class ComputationDocumentary(Scene):    def construct(self):        title = Text("What Is Computation?")         # With no duration, this longer built-in bed ends exactly with the body.        with self.play_sound(            "documentary/music/serious-ambient",            volume=0.25,            fade_in=1.0,            fade_out=2.0,        ):            self.play(Write(title), run_time=2)            self.wait(3)         with self.play_sound("documentary/transition/whoosh", volume=0.35):            self.play(title.animate.scale(0.85).to_edge(UP), run_time=0.6)         with self.play_sound("documentary/impact/low-cinematic-boom", volume=0.3):            self.play(FadeIn(Text("Universe → Information → Computation"))) 

A short built-in documentary cue is also kept as the runnable renderer fixturepackages/manim_renderer/tests/fixtures/documentary_sound_effect_scene.py and is used for real MP4 audio verification.

Verified original sources

Every source below is the original uploader's download and license page. Each is marked CC0 1.0: commercial use, editing, looping, layering, remixing, worldwide distribution, and monetized-video use are permitted; attribution and payment are not required. Formats and byte sizes describe the exact pinned downloads used by the generator, not an unrelated preview.

Sound ID Taxonomy

There are two intentional naming layers. Canonical IDs retain the pinned source-pack namespace, making every upstream file addressable without a handwritten Python subset. Semantic aliases provide stable intent-based names for generated scene code.

from manim import * print(sound_effect_count())       # 869 source assetsprint(len(sound_effect_names()))  # 899 names, including stable aliases for name in sound_effect_names():    if name.startswith("interface/"):        print(name) 

Combine All Three

Contexts may be nested. Each records its own start time and timing rule; the final mixer aligns all executed cues against the scene timeline.

from manim import * class ThreeAudioChannelsScene(VoiceoverScene):    def construct(self):        self.set_voiceovers({            "intro": {                "text": "Three independent audio channels, one final video.",                "mp3": "/data/audio/narration.ogg",            },        })        title = Text("Three audio channels")         with self.play_background_audio_block(            "/data/audio/theme.ogg",            volume=0.2,            fade_out=0.15,            loop=True,        ):            with self.play_voiceover_block("intro"):                with self.play_sound("ui/confirm", volume=0.7):                    self.play(Write(title), run_time=0.6)             with self.play_sound("impact/light"):                self.play(title.animate.scale(1.1), run_time=0.4) 

This exact pattern is kept as the runnable renderer fixturepackages/manim_renderer/tests/fixtures/three_audio_channels_scene.py. Replace its two workspace paths with uploaded narration and background files; bundled effect IDs need no upload.

Security, Packaging, And Performance

The generated catalog currently contains 266.7 MiB of original audio. Runtime compilation performs no network request. Archive checksums, direct-file checksums, per-file checksums, durations, count, aliases, and file existence are verified while generating the image.

Complete Sound Catalog

Filter by ID, source, or alias target. Click any body cell to copy that row's ID. The bounded table scrolls independently and keeps its header visible. It is generated at static-doc build time from the same checked manifest used by the Python renderer, so documentation and the container catalog stay aligned.