Refs
A video creation reference tool featuring a structured shot-by-shot analysis library of more than 3,000 films. Each film is deconstructed into 15 structural blueprints and recipe-style prompts. The library can be integrated into AI coding environments such as Claude Code, Codex, and Cursor via a single command, enabling AI systems to generate video scripts and storyboard plans with stronger cinematic language.
Timeline (last 90 days)
Refs is a tool that provides Claude with video reference materials; it connects to AI coding environments such as Claude Code, Codex, or Cursor via a single command and offers a structured database of cinematic blueprints.
Refs received 85 upvotes on Product Hunt, ranked 8th, and was categorized under three areas: developer tools, artificial intelligence, and video.
At the core of Refs is a material library containing more than 3,000 films.
Refs measures and analyzes each film shot by shot, organizing the results into a 15-plate blueprint and pairing them with ready-to-use recipe-style prompts.
Refs is essentially an implementation of the RAG approach in a vertical content domain, injecting cinematic blueprints as retrieved context at inference time rather than using them to train the model.
Refs most likely exposes its film blueprint database as an externally callable tool for AI via MCP (Model Context Protocol) or a similar tool-calling protocol.
The specific sources of Refs' films, how copyright is handled, the actual accuracy of the blueprints, and whether the generated results can reach a cinematic level have not yet been verified.
All Facts (7)
Refs is a tool that provides Claude with video reference materials; it connects to AI coding environments such as Claude Code, Codex, or Cursor via a single command and offers a structured database of cinematic blueprints.
50%UnverifiedRefs received 85 upvotes on Product Hunt, ranked 8th, and was categorized under three areas: developer tools, artificial intelligence, and video.
50%UnverifiedAt the core of Refs is a material library containing more than 3,000 films.
50%UnverifiedRefs measures and analyzes each film shot by shot, organizing the results into a 15-plate blueprint and pairing them with ready-to-use recipe-style prompts.
50%UnverifiedRefs is essentially an implementation of the RAG approach in a vertical content domain, injecting cinematic blueprints as retrieved context at inference time rather than using them to train the model.
50%UnverifiedRefs most likely exposes its film blueprint database as an externally callable tool for AI via MCP (Model Context Protocol) or a similar tool-calling protocol.
50%UnverifiedThe specific sources of Refs' films, how copyright is handled, the actual accuracy of the blueprints, and whether the generated results can reach a cinematic level have not yet been verified.
50%