What problem does it solve? Turning a script, article, or recorded narration into a finished, publishable video normally requires a desktop editor, a browser-based renderer, and hours of manual timeline work. This Skill automates the entire release pipeline inside a cloud sandbox: it derives scene timing from subtitle cues, renders frames with Pillow, encodes with ffmpeg, and produces the full publishing bundle — video, covers, platform copy, blog post, and social post. ## Core Features & Use Cases - Narration-driven scene planning: Parses an SRT file and groups cues into ~30-second scenes so picture changes land exactly on the words that motivate them, eliminating audio-video drift. - Cloud-native rendering: Composites 1080p scene stills with Pillow and encodes with ffmpeg using memory-safe crop-window motion, designed for a 2 vCPU / 2 GB sandbox with no browser. - Complete publishing bundle: Produces five-ratio covers (16:9, 16:10, 4:3, 3:4, 9:16), YouTube and Bilibili descriptions with chapters, a companion blog post, and a social post. - Automated verification: Checks that each scene appears when the narration says it should, that no two scenes share artwork, and that no frame is blank. - Use Case: A creator records a 7-minute narration for a faceless explainer, provides the MP3 and SRT, and receives a verified 1080p video, five thumbnails, and ready-to-paste platform copy without opening any editing software. ## Quick Start Turn my narration audio and subtitle file into a finished explainer video with covers and YouTube description copy.
