manim-video

Use Manim Community Edition to write one Python class per scene, render each scene with the manim CLI, then stitch clips with ffmpeg. This skill automates the full pipeline from narrative planning through final MP4 output.

You need Python 3.10+, Manim Community Edition v0.20+ installed via pip, a LaTeX distribution such as texlive or MacTeX for equation rendering, and ffmpeg for video stitching and audio muxing.

Yes, the manim-voiceover plugin integrates TTS services like ElevenLabs, Azure, or Google TTS directly into scene code and auto-syncs animation duration to narration length. Alternatively, you can mux audio manually with ffmpeg.

Manim's Pango renderer produces poor kerning with proportional fonts. Use monospace fonts such as Menlo for all Text objects, and preview text-heavy scenes at -qm quality since 480p draft renders hide readability issues.

Text objects are Mobjects, not VMobjects, so they cannot be added to a VGroup on Manim CE v0.20+. Use Group for mixed collections containing Text, and reserve VGroup for shapes and MathTex only.

Avoid Manim for single static diagrams, dense reference tables, or concepts better explained by pointing at one labeled figure. Animation adds cognitive load, so use it only when a sequence unfolds over time or spatial relationships change.