scenepilotLLM/scenepilot
ScenePilot turns your AI coding assistant into a full video production studio, open source
ScenePilot is an open-source system that lets AI coding assistants like Claude Code or Cursor handle an entire video production — research, scripting, asset generation, editing, and final rendering — from a single plain-language request. Instead of a hardcoded orchestrator, the AI agent itself reads YAML pipeline manifests and Markdown skill files to decide what to do at each stage. It can produce real videos with zero paid API keys using free TTS and free stock/archival footage, or scale up with paid generation providers for higher production value.
What it does
- A documentary montage pipeline builds a searchable corpus from free open archives like Archive.org, NASA, and Wikimedia Commons, retrieving actual motion footage rather than just animating still images.
- Every production follows a fixed seven-stage flow — research, proposal, script, scene plan, assets, edit, compose — with a dedicated Markdown 'director skill' guiding the agent through each stage.
- Provider choices for video, image, TTS, and music generation are scored across seven weighted dimensions (task fit, output quality, control, reliability, cost efficiency, latency, continuity), and the winning choice with its reasoning is logged.
- Before rendering, the system validates that the delivery promise is met and screens for 'slideshow risk'; after rendering, it runs ffprobe checks, frame sampling, and audio analysis, and only presents the video if it passes.
- Built-in budget governance sets a default total cap of $10 and pauses for approval on actions above a $0.50 threshold to prevent unexpected spending.
Why it matters
It removes the need to stitch together multiple separate AI tools manually for video production, and shows that real, footage-based videos are achievable using only free and open-source components. Its auditable, logged decision trail also offers a concrete example for anyone trying to verify how and why an AI-generated video ended up the way it did.
Terms in this repo
- TTS · Text-to-speech, technology that converts written text into spoken audio
- CLIP · An AI model that understands images and text together, enabling semantic search
- ffprobe · An FFmpeg tool used to inspect video and audio file properties
- Remotion · A rendering engine that programmatically builds video using React code
- YAML manifest · A configuration file that defines a pipeline's stages and settings
Repository description (English)
🔥 Turn any URL into clean, structured, LLM-ready content. The open-source web fetching & extraction API.
Open on GitHubTrending repos
- openai/codexOpenAI's coding agent, now runnable straight from your terminal
- cordiverse/cordisA TypeScript framework for snapping feature pieces on and off whenever you need them
- ripienaar/free-for-devA single repo lists every free-tier cloud, API and collaboration tool developers actually need
- Wei-Shaw/sub2apiA relay server that lets one Claude/OpenAI/Gemini/Grok subscription be shared by many users
- multica-ai/andrej-karpathy-skillsA single instruction file that stops AI coding assistants from making silent bad guesses
- eneskirca/nodetermA workspace that puts scattered terminal tabs and AI coding agents on one draggable map
- affaan-m/ECCA toolkit that teaches AI coding agents to work like disciplined engineers
- n8n-io/n8nn8n lets you build automation workflows and AI agents by dragging nodes on a canvas, with code as a fallback
Latest from METAL LAB
- Inherent's 27B Model Faraday Outperforms Anthropic and OpenAI
- LinkedIn's AI slop report button passes 1 million clicks
- NVIDIA pays startup Poolside $6 billion to license its AI model
- NVIDIA Swaps Harness, Lifts AI Agent Score from 30% to 100%
- Netflix Pilots In-House Language Model GenRec in Recommendation Engine