Skip to content

CLAUDE.md

Latest commit

 

History

History
123 lines (97 loc) · 6.25 KB

CLAUDE.md

File metadata and controls

123 lines (97 loc) · 6.25 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

This is the 3Blue1Brown video creation repository containing the Python code used to generate the mathematical animations and visualizations for the 3Blue1Brown YouTube channel. The project is built on top of the Manim animation library (specifically the 3b1b version, not the community edition).

Key Commands

Running Scenes

  • manimgl <file_name> <scene_name> - Render a scene to video
  • manimgl <file_name> <scene_name> -se <line_number> - Drop into interactive mode at a specific line (like a debugger)
  • manimgl <file_name> <scene_name> -p - Preview the scene without rendering to file

Interactive Development

  • checkpoint_paste() - In interactive mode, run code from clipboard with state management
  • checkpoint_paste(skip=True) - Run code without animation (zero runtime)
  • checkpoint_paste(record=True) - Record animations while running code

Staging Scenes

  • python stage_scenes.py <module_name> - Stage rendered scenes in order for a video module

Architecture

Directory Structure

  • _YYYY/ - Videos organized by year (e.g., _2025/, _2024/)
  • custom/ - Custom Manim extensions and reusable components
  • once_useful_constructs/ - Legacy utility classes and functions
  • outside_videos/ - Content for external collaborations and one-offs
  • sublime_custom_commands/ - Sublime Text editor integration

Core Files

  • manim_imports_ext.py - Universal import file that imports all Manim components plus custom extensions
  • custom_config.yml - Manim configuration with custom paths and rendering settings
  • playground.py - Sandbox file for testing and experimentation

Video Project Structure

Each video project typically contains:

  • Main scene files (e.g., main.py, part1.py, part2.py)
  • supplements.py - Additional scenes and helper functions
  • Helper modules for specific mathematical concepts

Custom Components

  • custom/characters/ - Pi creature animations and scenes
  • custom/backdrops.py - Background elements and visual themes
  • custom/drawings.py - Custom drawing utilities
  • custom/end_screen.py - Standard end screen components

Development Workflow

Scene Development

  1. Create scene classes inheriting from InteractiveScene or Scene
  2. Use manimgl with -se flag to develop interactively
  3. Use checkpoint_paste() to iterate on animation code
  4. Preview with -p flag before final rendering

File Organization

  • Start with from manim_imports_ext import * for all video files
  • Organize scenes chronologically within files
  • Use descriptive class names that match the video content
  • Group related scenes in the same file

Configuration

  • Camera resolution: 4K (3840x2160) at 30fps
  • Custom fonts and LaTeX configuration in custom_config.yml
  • Dropbox integration for asset management and video output

Code Patterns

Scene Classes

  • InteractiveScene - Base class for most scenes with interactive development support
  • PiCreatureScene - Scenes featuring the Pi creature character
  • TeacherStudentsScene - For Pi creature classroom interactions
  • Custom scene classes for specific mathematical contexts

Animation Patterns

  • Use self.play() for animations
  • self.wait() for pauses
  • self.add() for static elements
  • Color constants: BLUE, YELLOW, RED, etc.
  • Mathematical typesetting with Tex() (use this instead of MathTex() which is from ManimCommunity)
  • Format LaTeX strings with raw strings: Tex(R"\pi"), Tex(R"\frac{1}{2}"), etc.
  • Use lag_ratio parameter for staggered animations: FadeIn(objects, lag_ratio=0.1)
  • Common sequence: self.play(Write(equation)) followed by self.wait()

Mathematical Objects

  • NumberPlane and ComplexPlane for coordinate systems
  • ParametricCurve for mathematical curves
  • VGroup for grouping related objects
  • When creating multiple similar objects (e.g., multiple NumberLines), create them within a Group or VGroup to avoid code duplication
  • Custom mathematical visualization classes in once_useful_constructs/

Code Organization

  • Place configuration constants at the top of files
  • Use class attributes for scene-specific configuration: initial_positions = [10.5, 8]
  • Helper methods use get_* naming pattern: get_spring(), get_mass()
  • Access frame with frame = self.frame for camera operations

Color and Styling

  • Use text-to-color mapping for mathematical expressions: Tex(formula, t2c={"x": BLUE, "y": RED})
  • Consistent color schemes: earth=BLUE, sun=YELLOW, mathematical variables get specific colors
  • Set properties with chained methods: object.set_stroke(color, width).set_fill(color, opacity)

Updaters and Animation Control

  • Use updaters for dynamic positioning: object.add_updater(lambda m: m.move_to(target.get_center()))
  • Alternative syntax: object.f_always.move_to(reference.get_center)
  • Use LaggedStart(*animations, lag_ratio=0.1) for sequential overlapping animations

Python Code Style

  • Do not include indentation spaces on blank lines
  • Keep blank lines completely empty (no whitespace)
  • Prefer existing utility functions over one-off index math. For example, use color_gradient(colors, n) to produce a list of interpolated colors rather than calling interpolate_color manually inside a loop with i / (n - 1) arithmetic.
  • Construct VGroups declaratively when possible. If the full meaning of a group can be expressed in one expression, prefer VGroup(*[...]) or VGroup(... for ...) over a for loop that progressively calls .add(). Reserve the progressive style for cases where each iteration has non-trivial side effects.
  • Use existing layout methods (.arrange(), .next_to(), .move_to()) instead of manual in-place coordinate logic. A small amount of redundant movement is preferable to bespoke positioning arithmetic that is harder to read and easier to get wrong.

Notes

  • This repository uses the 3b1b version of Manim, not ManimCommunity
  • The project integrates with Sublime Text for enhanced development workflow
  • Video assets are managed through Dropbox with custom path configurations
  • Each year's videos are self-contained in their respective directories
  • No formal testing framework - scenes are tested through visual preview and rendering