The 3Blue1Brown YouTube channel built a visual language for math that millions of people now associate with "good math explanation." Smooth dark background, glowing white equations that morph step by step, a few accent colors used sparingly, and a pace that lets you absorb each move. If you have ever watched the linear algebra series or the calculus playlist, you know the look.
You also probably know the reason most people never make videos in that style: it requires Python. Grant Sanderson built Manim, a custom animation library, specifically for these videos. The community-maintained fork is what almost everyone uses today, and it is still a Python library. You write code. You render. You debug.
This post is for people who want the look without the climb. We cover what defines the aesthetic, why slide tools cannot fake it, what your options are, and a worked example using AI prompting. The honest version, including where it breaks.
What defines the 3Blue1Brown aesthetic
When people say "3Blue1Brown style," they usually mean a specific bundle of choices:
Dark background. Not pure black. A muted dark navy that lets glowing text feel almost luminous against it. The default Manim background color (#000000 adjusted to roughly #0E1525 in the channel's videos) reduces eye strain in long-form watching.
Off-white text and equations. Pure white is harsh. Slightly warm off-white reads as written, not displayed. LaTeX equations in this color, with smooth font rendering, look closer to chalk on a clean board than to a slide deck.
Small palette of accent colors. Teal, yellow, red, and a soft blue. Used to mark "this is the part to look at." Every other element fades back. The discipline is what makes complex equations readable.
Gradual reveals and transforms. The signature move: an equation appears, a piece of it animates into a new form, the new form sits long enough to absorb. No flashy zooms. Camera moves are minimal and serve the math.
Mathematical typography. LaTeX rendered with a serif math font. Variables in italic. Operators kerned correctly. This is the part slide tools cannot match without an external math tool.
The combination is hard to copy by accident. Hard to copy on purpose, too, but for a different reason: each element is the result of a deliberate choice, and Manim makes those choices easier than other tools because the library is built for it.
Why slide tools and standard animation software fail
People often try Keynote, PowerPoint, or general animation tools first. The result is always close-but-wrong. Why:
- No real LaTeX. PowerPoint's equation editor renders text, not vector math glyphs. The kerning is off, the integral signs are blocky, the fractions look pasted in.
- No path morphing. A core 3Blue1Brown move is morphing one equation into another along the implicit transformation. Slide tools fade-out, fade-in. They cannot interpolate the strokes of
x^2 + y^2into(x + y)^2 - 2xy. - Manual timing. Every reveal in a slide deck is a hand-tuned click event. Manim describes timing in code, so you can specify "wait 0.5 seconds, then transform" and have it consistent across scenes.
- Color discipline is hard. In a GUI, it is easy to add too many colors. In code, you import a palette once and reuse it.
Tools like Veed or Animaker are great for talking-head explainers, but they do not produce the math typography or the smooth morphs.
The Manim option (high effort)
If you have a few weekends and basic Python, learning Manim is worth it. The community edition's tutorial walks you through your first scene in about 30 minutes. Building a polished one-minute video takes longer: 4 to 8 hours for a beginner, 1 to 2 for someone fluent.
The cost is real. You need:
- Python 3.9 or newer.
- A working LaTeX install (MacTeX on macOS, MikTeX on Windows, TeXLive on Linux).
- ffmpeg.
- Patience for the Manim API: scenes, mobjects, animations, configuration files.
Once installed, the iteration loop is fast. Edit code, run manim -pql scene.py, watch the preview. The community has a large catalog of example scenes you can adapt.
For people willing to invest, this is the highest-control path. The full 3Blue1Brown look is achievable. You will spend most of your time on layout and timing, not on syntax.
The AI-assisted route
For people not willing to invest, AI tools that generate Manim code from prompts are the alternative. Madio is one. The flow is:
- Open /create.
- Type a prompt: "show the Pythagorean theorem visually, with a right triangle, squares on each side, and a morph showing a^2 + b^2 = c^2."
- Wait 60 to 180 seconds.
- Get an MP4. On the Starter plan and above, also get the Python source.
Under the hood, Google Gemini 3 Flash or Pro Thinking produces Manim code, the system runs it through Manim Community v0.18.1 in a sandbox, and returns the result. There is a retry loop for syntax errors (the most common failure mode is Tex versus MathTex confusion, which we patch automatically).
The output looks like 3Blue1Brown for simple prompts. The default palette is close to the channel's. Equations render with LaTeX. Transformations morph rather than cut. For most users, the result is "this looks like a 3Blue1Brown video for this one moment."
Worked example: Pythagorean theorem
Let me walk through one prompt and what comes out.
Prompt:
Show the Pythagorean theorem. Start with a right triangle with sides labeled a, b, c where c is the hypotenuse. Add a square on each side, labeled a squared, b squared, c squared. Then animate the two smaller squares moving together to form the larger square. End with the equation a squared plus b squared equals c squared.
What Madio produces (roughly):
A 30-second clip. Dark background. The triangle appears stroke by stroke. Labels fade in next to each side. The three squares grow from each side outward. Each square is filled with a soft accent color. The two smaller squares animate by translating and reshaping until they together cover the larger square. The final equation glows in the center.
For a 30-second free-tier render, the output is recognizable as the genre. Not pixel-identical to a Grant Sanderson video. The pacing is faster than his (his videos breathe more). The colors are close but not exact. The camera does not move; his often does.
If you take the downloaded .py (Starter plan) and tweak self.wait(1) to self.wait(2) in a few places, the pacing improves dramatically. If you swap the default colors for #3FBDF6 and #FCD34D, you match the channel more closely. Five minutes of editing for noticeably better output.
What works and what does not
Works well:
- Single-concept geometry (triangle, circle, polygon proofs).
- Equation morphs (factoring, expanding, substitution).
- 2D function plots and basic transformations.
- Counting and combinatorics with discrete elements.
- Probability tree visualizations.
Does not work as well:
- Complex 3D scenes (Manim's 3D is harder for LLMs to reason about).
- Long videos with multiple narrative beats.
- Custom mobjects that are not in Manim's standard library.
- Synchronized voice and animation timing (Madio's narration on Pro is automated, not artisan).
- Hand-drawn or sketchy aesthetics (this is a vector library, not a paint program).
A useful mental model: AI tools are a draft generator. They get you 70 percent of the way to a 3Blue1Brown look. The last 30 percent is hand work, either by editing the prompt and re-rendering, or by downloading the source and tweaking it.
Practical workflow
Here is how I would approach a 3Blue1Brown-style video as a non-coder:
- Sketch your concept on paper. What is the one idea? What is the visual that conveys it? If you cannot describe it in two sentences, the AI will not handle it well either.
- Write a specific prompt. Name the objects, the colors if you care, the transformation. Specific prompts succeed more often than open-ended ones. See our prompting patterns post for templates.
- Render on the Free tier first. Five credits gets you five experiments. Use them to find the prompt that works, then upgrade only if you need length, quality, or downloads.
- Browse the gallery to see other people's prompts and steal the pattern.
- If you need polish, upgrade to Starter. Get the
.py, edit timings and colors, re-render with Manim CE locally if you want. - For consistent series, learn just enough Manim to refactor. A few hours with the community docs gets you to "I can edit AI-generated code." That is the sweet spot.
On Grant Sanderson and the channel itself
Grant Sanderson built Manim as a tool for himself. The fact that he open-sourced it (and that the community maintains a robust fork) is what makes any of this possible. None of the AI tools, Madio included, would exist without his work.
If you watch his videos, you see what hand-crafted artistic direction looks like. The pacing, the analogies, the music, the subtle camera moves. AI does not match this. If you want to replicate his channel, you will need to learn his craft. AI gets you the visual language. The narrative voice is yours to build.
Try it
You do not need a credit card. Start with five free credits and a single prompt. Browse /templates for prompts on common topics: linear algebra, calculus, geometry, probability. If you want to compare AI-prompt output to other animation tools, see /compare/madio-vs-veed.
For a deeper look at what AI tools currently do in this category, our roundup of text-to-Manim AI tools in 2026 compares hand-rolled GPT prompting, open-source projects, and hosted offerings. For the technical side of how a prompt becomes an MP4, see the AI animation pipeline post.
Final word
The aesthetic that 3Blue1Brown popularized is reachable today without writing Python. The AI tools are not perfect. They do not match Grant Sanderson's editorial polish, and they probably never will, because that polish comes from a human making careful choices. But for the visual language (dark background, glowing math, smooth transforms), an AI tool gets you there in 90 seconds, free of charge for the first few attempts.
If your goal is "I want a video that explains this one idea, today, in a style that looks at home next to a 3Blue1Brown video," that is now achievable. If your goal is "I want to make a full series like his," set aside time to learn the craft. There is no shortcut for that part.
For now, write a prompt, hit render, and see what comes out.