Routes your video request through RunComfy's CLI to the right model: HappyHorse 1.0 if you want in-pass audio, Veo 3-1 for physics-accurate product spins, Wan 2-7 when you need lip-sync to a specific voiceover file, Kling 3.0 4K for final delivery, Seedance v2 for multi-reference cinematic work. Covers text-to-video, image-to-video, and video extension. The model catalog is deep enough that picking the wrong one wastes time and money, so the skill ships prompting patterns and intent heuristics for each. Helpful if you're doing ad creative or social clips at volume and don't want to memorize which ByteDance tier does what.
npx skills add https://github.com/doany-ai/skills --skill ai-video-generation