What problem does it solve? Creating cinematic AI video with consistent characters, native audio, and the right quality tier requires navigating six different Kling 3.0 endpoints, pricing tiers, and prompting conventions. This Skill routes requests to the correct RunComfy endpoint and handles CLI invocation, polling, and result download automatically. ## Core Features & Use Cases - Six endpoint coverage: Routes across three rendering tiers (Standard, Pro, 4K) and two modes (text-to-video, image-to-video) based on budget and quality needs. - Multi-shot generation with audio: Supports numbered shot sequences with identity preservation and one-pass synchronized dialogue or ambient sound via generate_audio. - Tier and cost guidance: Provides per-second pricing, duration limits (3-15s), and aspect ratio constraints so you can validate cheaply on Standard before committing to 4K. - Use Case: A marketer needs a 10-second 4K brand film with a consistent character across three shots and background music. The Skill selects the kling/kling-3.0/4k/text-to-video endpoint, builds the multi-shot prompt, runs the RunComfy CLI, and downloads the finished MP4. ## Quick Start Ask the agent to generate a 5-second Kling 3.0 video of a barista pulling an espresso shot in 9:16 vertical format using the Standard tier.