Introducing OpenVideoKit

Introducing OpenVideoKit

OpenVideoKit is a scene-based, AI-first video editor. You edit slides in a browser UI, a Python server stamps values into self-contained GSAP compositions, <hyperframes-player> renders them live, and edge-tts generates the voiceover. Export to MP4 via HyperFrames.

Watch the demo

Template showcase

A walkthrough of the editor, the template catalog, and an end-to-end render.

OpenVideoKit template showcase

Text-to-Speech engine — 7 languages in 44 seconds

edge-tts voiceover showcase, cycling through seven locales on the same slide:

Global Voices demo — edge-tts voiceover across 7 languages

Cycle: 🇺🇸 Ava Multilingual → 🇫🇷 Vivienne → 🇩🇪 Florian → 🇯🇵 Nanami → 🇻🇳 Hoài My → 🇨🇳 Xiaoxiao → 🇧🇷 Thalita → 🇺🇸 Ava Multilingual.

Browse the catalog

The showcase above draws from the 25 curated slide layouts that ship with OpenVideoKit — titles, charts, kinetic typography, SVG diagrams, and more. Each one is a bare <template> that you, the AI copilot, or any MCP-connected agent can apply to a slide in a single call. 👉 Browse the templates gallery → for live video previews of every layout, field defaults, and the HTML source.

A built-in AI copilot

A LangGraph agent reads your project with read-only tools and proposes edits as the same EditOps a human dispatches. You accept or reject each proposal; undo/redo works uniformly. “AI flow == human flow” isn’t a slogan — it’s how the EditBus is wired.

  • One proposal per turn, then STOP — no autonomous editing.
  • Accept/Reject receipts are fed back to the model so it learns from “no”.
  • HTML lint, voice-id validation, and prompt-injection guards gate every edit.
The editor showing the AI prompt that generated the demo video on the left

The editor showing the AI prompt that generated the demo video above.

MCP for any agent

The MCP server is enabled by default at /mcp/. Claude Code, Cursor, Gemini CLI, and Goose connect directly to the same 28-tool API the copilot uses — no file scraping, mutations go through the official Python API.

Bring your own key

Works with OpenAI, OpenRouter, Ollama, vLLM, and LM Studio. Inference runs on your own machine; the cloud control plane never proxies and never charges. Local / BYO-key AI is free forever.

WYSIWYG by construction

Preview and MP4 export share one HyperFrames engine. “It looked different when I exported” is structurally impossible.

Ready to dive in? Browse the templates gallery →, then grab the source on GitHub.