What problem does it solve? Turning a topic, README, or screen recording into a finished narrated video normally requires stitching together scripting, TTS, image generation, captions, and a renderer by hand. This Skill automates that entire pipeline through a single CLI with deterministic, reproducible run directories. ## Core Features & Use Cases - Three generation modes: shorts/reels (9:16) from a topic, explainers (16:9) from README/code/data, and demos from screen-capture files or supervised headed-browser web capture of any URL. - Key-optional 3-tier router: every capability (voice, visuals, captions, compositor) has a real provider path and a key-free fallback (oma-voice, oma-image, oma-slide, Remotion), so it works with zero API keys and upgrades automatically when keys like PEXELS_API_KEY are present. - Deterministic asset bus: each run produces script.json, timing.json, render-spec.json, captions, the mp4, and a manifest.json with hashes, providers, and cost breakdown; runs can be re-rendered byte-stably from render-spec.json. - Use Case: Ask for a 30-second vertical short about a product feature; the Skill infers the mode, generates the script, synthesizes narration, creates Ken Burns stills, adds TikTok-style captions, and renders a finished mp4 with a full reproducibility manifest. ## Quick Start Ask the agent to make a 30-second short about your topic, for example: create a 30s vertical short explaining this project's README with TikTok captions.
