marianoperezbaldasare-maker/videocode

Videocode is a Python tool that turns coding videos into runnable code files. Developers often learn from video tutorials, but extracting the code shown on screen means pausing, rewinding, and manually retyping everything, which introduces typos and loses context. Videocode automates that step. You point it at a local video file or a YouTube link, and it produces a folder of working code. Internally it uses FFmpeg to break the video into scenes such as intro, coding, demo, and outro, a frame selector to pick the moments where code actually appears, Whisper to transcribe the narration for extra context, and a vision language model to read the code out of the selected frames. The extracted pieces are then assembled into a runnable project with dependencies resolved. Everything can run entirely on your own computer using Ollama, a tool for running AI models locally, so no API key is required and no data leaves your machine. For higher accuracy, you can instead connect a cloud backend such as Gemini, OpenAI, or Qwen. Videocode also works as an MCP server, meaning it can be called directly by Claude Code as a tool. MCP, or Model Context Protocol, is a standard way for AI assistants to use external tools, so once the server is running you can simply ask Claude Code to pull the code out of a video as part of a normal conversation. Beyond code extraction, the project includes a mode for finding bugs in screen recordings, an OCR mode for pulling text out of slides and dashboards, and a feature that tries to locate a tutorial's real GitHub repository from the video's metadata, so you do not have to retype code that is already published somewhere. The project is released under the MIT license and is aimed at developers who spend time learning from video content.