VisoCode

🙋‍♂️➡️🎞️ VisoCode turns questions into cinematic explanations—powered by AI agents, rendered with Manim.

---

🧠 Overview

VisoCode is an AI-powered explainer engine that transforms questions into expert-level animated videos using a multi-agent pipeline. Whether you're curious about medical AI, calculus or the code base in front of you, Visocode gives you not just an answer, but an explanation you can see.

---

🔄 How It Works

1. 🙋‍♂️ You ask a question
2. 🧠 ResearchAgent (powered by Kimi-K2) builds an expert explanation using:
   • 🔍 arXiv search
   • 🧮 math/calculation tools
   • 🖼️ icon generation
   • 🧾 structured metadata (sections, visuals, references)
3. 📤 The explanation and metadata are passed to a second agent...
4. 💻 CodeAgent (powered by Gemini) turns that explanation into:
   • 🎬 Writes frame-accurate Manim scene code
   • 📐 Converts math into LaTeX-rendered visuals
   • 🧩 Coordinates icons, timing, and transitions for clarity
5. 🎞️ Final Output: A polished Manim animation that visualizes your question with expert clarity.

---

🚀 Why VisoCode?

• ✨ Visual-first learning: Abstract concepts come to life through animation
• 🔗 Agent-powered pipeline: Separates research and rendering for modularity, traceability, and full control
• ⚡ Fast, automated output: From natural language to video in minutes
• 🧠 Research-grade insight: Built on real scientific search + analysis

---

💡 Who Is It For?

• 👩‍💻 Developers exploring unfamiliar systems or concepts
• 🎓 Students reviewing math, physics, CS, and more
• 👨‍🏫 Teachers designing custom visual explanations
• 📹 Content creators automating technical YouTube videos

---

VisoCode isn’t just a tool. It’s a thinking partner and video studio, all in one intelligent pipeline.

---

Installation

1. Windows 11 (TODO: support other OS by isolating Windows-specific actions in some microservice)
2. Git
3. Python >=3.11.6, <3.12
4. VapourSynth R65
5. uv
6. Manim
   • Install a LaTeX distribution as well
7. ffmpeg
   • Add executable to path
8. tex-fmt
   • Installation involves cargo, the Rust package manager
   • Make sure executable is in path
9. VS Code or a fork like Cursor
10. Clone the repo
11. Open in VS Code
12. Install recommended extensions in .vscode/extensions.json
13. Create and populate .env following .env.example

14. uv sync

15. Activate venv