ManimCat
Describe a math idea in plain language and ManimCat plans it, writes the Manim code, renders it, and hands back a video or image ready for a lesson. A planning card settles the question, the one-line answer, and the step structure first, and every card is editable before generation starts. Render failures go back to the model automatically, finished clips re-render from saved code, and an AI edit pass changes only what you ask for. Reference images can be uploaded or sketched on a built-in canvas with freehand strokes, shapes, text, and live function plots. Edge TTS narration is aligned to sentence-level cues, with soft subtitles or burned-in hard subtitles on download. The in-development Agent mode opens Manim Studio and Plot Studio, where a builder agent reads, writes, edits, and searches the Manim API with every tool call visible. Each pipeline role has an editable prompt template, and the interface offers English, Chinese, and light and dark themes. Bring any OpenAI-compatible key or set server-side model routes. Submitted code runs in a mandatory bubblewrap sandbox with no network access. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0-or-later licensed.
Deploy