Video Use (mcp-video edition)

Principle

1. mcp-video is the editing surface. All video operations go through its MCP tools (119 tools), Python client, or CLI. Never hand-write raw FFmpeg commands — use structured mcp-video tools instead.
2. LLM reasons from raw transcript + on-demand visuals. Derive cut candidates from speech boundaries and silence gaps. Drill into visuals only at decision points.
3. Ask → confirm → execute → iterate → persist. Never touch the cut until the user has confirmed the strategy in plain English.
4. Generalize. Do not assume what kind of video this is. Look at the material, ask the user, then edit.
5. Artistic freedom is the default. Every specific value, preset, font, color, duration in this document is a worked example — not a mandate. Make your own taste calls.
6. Invent freely. mcp-video supports trim, merge, resize, crop, rotate, overlays, subtitles, fades, effects, transitions, text, chroma-key, filters, speed changes, audio normalization, image analysis, and more. If the material calls for a technique not described here, mcp-video likely has a tool for it.
7. Verify your own output before showing it to the user. Use video-quality-check and video-release-checkpoint before publishing.

Hard Rules (production correctness — non-negotiable)