ai-video-modelslisted

Per-engine prompting guides for AI video models (Seedance 2.0, Veo 3.1, Kling). Genre-neutral — how to write a good prompt for each engine (reference-image token, prompt order, length, motion language, FLF, duration caps, style whitelist). Loaded BY creation skills (ugc-product-video, generic-video) once a model is chosen — NOT a standalone entry point.
Nagellabs/libi · ★ 2 · AI & Automation · score 73

Install: claude install-skill Nagellabs/libi

# AI Video Models — per-engine prompting guides Shared reference. A creation skill (`ugc-product-video`, `generic-video`) loads this once it has chosen a video model, to read how to prompt that specific engine. Scope boundary (do not duplicate across skills): - The **craft** (9-layer formula, clip-duration methodology, realism cues, negative lists) lives in `ugc-craft`. - The **mechanics** (provider discovery, `recommend_model` / `get_model_schema` / `get_pricing`, polling, import) live in `ai-asset-generation`. - **This skill is ONLY the per-engine prompt rules.** - Genre **use-case formulas** (e.g. the UGC product-hero / feature-walkthrough recipes) live in the calling creation skill (e.g. `ugc-product-video`), not here. ## Pick the guide for your chosen model - **Seedance 2.0** (strong default for most video) → [model-seedance-2](prompts/model-seedance-2.md) - **Veo 3.1** → [model-veo-3-1](prompts/model-veo-3-1.md) - **Kling** → [model-kling](prompts/model-kling.md) Always verify the model at runtime via `ai-asset-generation` (`recommend_model` → `get_model_schema` → `get_pricing`) before composing — availability, schemas, and pricing drift. Never trust a hardcoded model id as ground truth.