thumbnail-maker

<objective> Thumbnail generator. Input: title + optional face photo + optional style. Output: N variants at 16:9 (YouTube / blog / podcast-episode thumbnail). Distinct from `cover-maker`: - Aspect: ALWAYS 16:9 (no medium variation) - Aesthetic: bright contrast, large text, exaggerated facial expressions — "thumbnail look" - Variants vary in FACE PLACEMENT (left / right / center) more than in style Distinct from `flyer-maker`: - No event details (date / location) - Stronger emphasis on the FACE if photo provided - Different typography conventions (bigger, often colored, with stroke / shadow) This skill does NOT: - Generate YouTube banner art (different aspect — use `cover-maker --aspect 2560x1440` or `image-prompt --execute`) - Generate animated thumbnails — static images only - Source rights-cleared face references — provide your own - Optimize for click-through (subjective — picks best of 3-5 manually) </objective> ## ROLE Read title + optional face → pick model (text-friendly + ref-capable) → assemble per-variant prompts with face-placement + bold title composition → batch execute → save PNGs. ## PIPELINE 1. **Resolve title**: - Required: `--title` - 3-7 words is the sweet spot for thumbnails - High-energy phrasing works best ("How I Built X in 30 Days", "STOP DOING THIS") 2. **Resolve photo** (optional): - `--photo <path-or-url>` — face reference - Identity preserve via Nano Banana Pro (default) 3. **Resolve style** — see `references/composition-zon...