We made a 3blue1brown-style explainer in an afternoon

We've always wanted explanations the way 3blue1brown does them — the math drawn, not recited. It turns out the whole pipeline is now something you can just... run.

The video below is one we made. I'm learning to use a 3D printer, and I was curious how one actually gets calibrated along its three axes — and how a printer like a Bambu Lab calibrates itself. So that became the test topic.

What's in it: steps-per-mm as a one-parameter regression, frame skew as a shear matrix the firmware inverts, bed leveling as least-squares plane fitting plus bilinear interpolation, and Bambu's two party tricks — resonance hunting with input shaping, and pressure advance fitted by its own lidar. Every calibration step is the machine running an experiment on itself.

The recipe

Three ingredients, all public:

1. Manim — Grant Sanderson's animation engine, the same one behind the real videos (we used the community edition).
2. manim-skill — a Claude Code skill by Yusuke710 that turns a topic into a planned, scripted, rendered video: Plan → Code → Render → Iterate. Pair it with manimce-best-practices by Adithya S K.
3. ElevenLabs narration — each caption line becomes a voice clip; we ran it through Higgsfield's TTS with an ElevenLabs voice.

Install the skill, ask for a topic, get a video.

The one non-obvious trick

Narration never fits animation timing on the first try — ours ran two minutes longer than the video. The fix is to do what 3blue1brown actually does: retime the video to the voice, not the voice to the video. Measure each clip, stretch each scene's holds to fit, re-render. Animations keep their speed; the frame simply holds while the narrator finishes the thought. Sync error ends up under a second across five and a half minutes.

Try it yourself

Nothing in this pipeline is custom to us. Paste the prompt below into Claude Code (or any coding agent), swap in your own topic, and it will set up the whole environment and make you a video:

Set up my machine to create 3blue1brown-style explainer videos, then make one for me.

1. Install the skills:
   - Clone https://github.com/Yusuke710/manim-skill and copy skills/manim-skill into ~/.claude/skills/
   - Clone https://github.com/adithya-s-k/manim_skill and copy skills/manimce-best-practices into ~/.claude/skills/
2. Install Manim Community Edition and its dependencies (on macOS: brew install cairo pkg-config ffmpeg, then uv tool install manim). Verify manim --version works.
3. If LaTeX is not installed, use Text with unicode math symbols instead of MathTex, and keep axis numbers off.
4. Using the manim-skill workflow (Plan -> Code -> Render -> Iterate), make a ~3 minute explainer video about: [YOUR TOPIC HERE]
5. Add narration: treat each on-screen caption as one cue, synthesize each cue with a natural TTS voice, then retime the VIDEO to the audio - extend each scene's holds so every clip fits its cue window - and mux the track in. Never speed up the audio to fit the video.

That's it. The gap between "curious about something" and "watchable explainer about it" is now one afternoon.