cinematic-director

Describe photographic decisions instead of objects: state the intent, subject hierarchy with exact frame percentages, where the character's attention falls, lens and camera position, and restraint clauses. This Skill automates that process through a simulated photography crew meeting and seven decision modules.

Remove quality-bait keywords like 8K, masterpiece, HDR, and ultra detailed, which push models toward CG promotional art. Use low contrast, restrained highlights, motivated lighting, limited 3-5 color palettes, and negative prompts blocking CGI, game render, and poster composition terms.

Yes. For SDXL and ComfyUI it outputs a full four-category negative prompt; for Midjourney it condenses negatives into a --no parameter list; for models without negative support like GPT Image or Gemini it folds avoidance rules into 2-4 natural-language sentences at the end of the positive prompt.

A negative prompt lists visual traits the model should avoid. This Skill organizes it into four classes: AI-artifacts (HDR, plastic skin), non-cinematic composition (centered, hero pose), lighting problems (fake rim light, blown highlights), and color issues (oversaturated, neon), plus genre-specific and scene-consistency blocks.

Do not use it for technical consistency assets such as multi-object reference sheets, three-view prop boards, or evenly lit catalog-style images. Those require neutral backgrounds and uniform lighting, which this Skill explicitly excludes; its still-life mode only covers single-object atmospheric shots.