What problem does it solve? Creating AI-generated video with synchronized audio and consistent characters across shots requires knowing the right model, parameters, and prompting techniques. This Skill routes text-to-video requests to HappyHorse 1.0 on the RunComfy Model API with the correct schema, prompting patterns, and model-selection guidance. ## Core Features & Use Cases - Text-to-video generation: Submits prompts to the happyhorse/happyhorse-1-0/text-to-video endpoint with control over aspect ratio, resolution (720P/1080P), duration (3-15s), seed, and watermark. - Multi-shot character consistency: Documents prompting patterns that restate character anchors across shots so identity and wardrobe persist between scenes. - In-pass synchronized audio: Supports audio direction in prompts (dialogue, ambient sound, Foley) generated in the same pass as the video. - Model routing guidance: Compares HappyHorse 1.0 against Wan 2.7, Seedance 2.0 Pro, LTX 2, and Kling Video O1 so requests land on the right model. - Use Case: A marketer needs a 9:16 vertical ad with a consistent spokesperson and voiceover. The Skill builds the multi-shot prompt, submits it via runcomfy run, polls until completion, and downloads the 1080p result to a local directory. ## Quick Start Ask the agent to generate a video with HappyHorse, for example: "Use HappyHorse to generate an 8-second 9:16 video of a barista pulling an espresso shot with cafe ambience audio, saved to my Downloads folder."
