What problem does it solve? Turning a single topic into a complete multi-scene video requires manually coordinating text-to-image generation, image-to-video rendering, polling, downloading, and concatenation, with failures at any step breaking the whole run. This Skill automates that entire pipeline using a Finite Execution Graph with a per-scene state machine, quality-gate hooks, and automatic retry/skip logic. ## Core Features & Use Cases - FEG + State Machine Pipeline: Each scene flows through IMAGE_GEN, VIDEO_GEN, POLLING, DOWNLOADING, and COMPLETE states with up to 3 automatic retries and graceful skipping of failed scenes. - Quality-Gate Hooks: Pre-flight checks (API key symlink, httpx, ffmpeg, disk space), post-image/post-video URL validation, and pre-concat resolution checks via ffprobe. - Character Consistency Mode: idol-video.py locks a reference image across all scenes so the character appearance stays consistent, with negative prompts preventing identity drift. - Use Case: Ask for a 5-scene animated short about a topic; the pipeline generates images with Agnes Image 2.1 Flash, renders clips with Agnes Video V2.0, downloads each MP4, and concatenates them with ffmpeg into one final video. ## Quick Start Run the FEG pipeline script with a topic, scene count, and duration to automatically generate and concatenate a multi-scene video.
