vargHQ/sdk — reverse-engineered prompt
Reverse engineered prompt
Build me an open source TypeScript SDK for creating AI videos with JSX, kind of like React for video scenes. I want someone to describe clips, images, voiceovers, music, captions, transitions, talking heads, and product end cards in simple components, then render an mp4 locally or send it to a cloud render API.
It should use one varg API key and one gateway that can route to providers like Kling, Flux, ElevenLabs, Sora, Wan, and similar models. Please include caching so rerunning the same prompt does not regenerate paid assets. Add a simple CLI that can initialize a project, create a starter video file, and render it.
Include practical examples like image to video, voiceover with TikTok style captions, background music, and lipsync talking head. Make the developer experience clean for Cursor or Claude Code users, with docs and enough tests to show the main flows work. Look up current docs online if you need to.
Want more depth? Deep Reverse