live-sketch-motion

Use for a real photo, video, or approved realistic preview where a foreground marker visibly and progressively converts one still-moving subject into a colored-pencil sketch. Unlike animate-image (rel

SKILL.md

GitHub

Live Sketch Motion

Catalog

  • Display name: Live Sketch Motion
  • Categories: Video
  • Status: Ready
  • Route label: Marker-synchronized live-to-sketch conversion
  • Tone: blue
  • Short description: Animate a real scene while a visible marker progressively redraws one moving subject as a lively colored-pencil sketch.

What this skill does

  • Preserves source composition, identity, support surfaces, and plausible background micro-motion.
  • Specifies a natural marker hand, tip-synchronized reveal, and a three-layer white-border, black-line, hatched-fill sketch.
  • Prevents frozen subjects, pre-reveals, sticker pop-ins, and generic overlay effects.

How to use

Provide a photo, a motion reference, or a text scene; include target subject, ratio, duration, sketch style, and sound preference. Text-only scenes receive a realistic preview gate. Canvas-capable Agents may store plan and prompt documents there; otherwise they deliver them in chat or locally.

Example

把这张跑步小狗照片做成 7 秒横版视频:蓝色马克笔从右下进入,边画边把小狗变成彩铅涂鸦,小狗全程继续跑。

Workflow

  1. For text-only requests, create and approve a realistic source preview. For supplied media, lock composition, identity, subject path, support contact, and camera compatibility.
  2. Plan marker entry, natural grip, tip contact, upper-left-to-lower-right reveal, hand exit, and continuous subject motion.
  3. Send one approved clip to $generate-video with source roles explicit. Persist an operationKey, poll only the accepted receipt, and stop to reconcile unknown outcomes.
  4. Inspect the opening, reveal, and ending frames for no pre-reveal, no freeze, correct layer order, plausible occlusion, ratio, and audio policy.

User-facing output

  • Media type: One MP4 video and optional approved preview
  • Default quantity: 1
  • Content per video: One continuous marker-driven subject conversion
  • Default layout: Source ratio or 16:9
  • Model policy: Verified image route for text-only preview; catalog-selected video route for clip
  • Downstream use: Social and creator visual effects

Route

  • Upstream assets: Real photo, reference video, or approved preview
  • Downstream handoff: $combine-videos only for accepted clips