What problem does it solve? Choosing the right AI video model and writing a correct API call for it is tedious: each model (Kling, Veo, Seedance, Wan, HappyHorse, Hailuo, Dreamina) has different strengths, endpoints, and prompt conventions. This Skill routes a user's video request to the right RunComfy model and produces the exact runcomfy run invocation with documented prompting patterns. ## Core Features & Use Cases - Model routing by intent: Picks the right model for Arena-#1 quality, multi-shot character identity, in-pass audio, physics-accurate motion, lip-sync to a voiceover file, fastest iteration, or 4K final delivery. - Three generation modes: Text-to-video (t2v), image-to-video (i2v), and Veo's extend-video endpoint for continuing existing clips. - Ready-to-run CLI invocations: Ships schema tables, prompt tips, and exact runcomfy run <vendor>/<model>/<endpoint> commands for each route. - Use Case: A marketer asks for a 9:16 product spin with no background motion. The Skill selects Veo 3-1 i2v, supplies the product image URL and a physics-aware prompt, and runs the CLI to download the finished clip. ## Quick Start Ask the AI to generate a video from your prompt or still image, for example: generate an 8-second 16:9 video of a red kite tumbling across a windy beach at golden hour with ambient audio.
