Add a restrained depth wiggle to one image using editable, hand-authored depth regions. Manual rectangular and elliptical depth regions form a smooth displacement field. Inverse sampling shifts foreground and background by different amounts while a fixed crop protects the borders. PopNow can supply missing artwork, footage or sound after a quote; owned material and the local example work without a connection.

Uses PopNow for
Images · Music & voice
Needs
Python 3 · FFmpeg

The skill is free. Optional AI images, video and sound are quoted for your approval before generation. API-key calls use your separate API balance in USD; generation through ChatGPT or Claude connected via OAuth/MCP uses your website credits.

v1.0.0 · Updated Oct 25, 2026 · 3.5 MB · Use it for personal and client work; please don't resell or share the package.

What you get

  1. An 8-second 1280 × 720 H.264 MP4 made by the local algorithm

  2. Manual rectangular and elliptical depth regions form a smooth displacement field. Inverse sampling shifts foreground and background by different amounts while a fixed crop protects the borders.

  3. Editable JSON, a six-frame contact sheet and a receipt with actual checks

Good for

  • Paper dioramas, quiet photo loops and illustrated scene studies
  • Original creative work using owned or licensed material

Not for

  • Automatic depth inference, true 3D reconstruction or newly revealed hidden surfaces
  • Publishing without watching the rendered result

How to use it

You say

  • Use the PopNow skill "Depth Wiggle Parallax" (depth-wiggle-parallax) for this: Add a restrained depth wiggle to one image using editable, hand-authored depth regions. Use the bundled project, render the example and inspect the full result before delivery.

Then try

  • Change amplitude from 25 to 12 for a calmer depth wiggle.

  • Use my licensed music; keep source audio discarded.

What to have ready

  • One layered scene and a rough, manually authored foreground-to-background plan.
  • Python 3.10+, FFmpeg/ffprobe, Pillow and NumPy

Similar skills

View all →

Ratings

No ratings yet. Installed it? Be the first to rate it.

Rate this skill

Questions

Does it infer depth with AI?

No. You author approximate regions and their relative depth. The preview is a 2.5D effect, not evidence of recovered scene geometry.

Can it reveal what was behind an object?

No. Dense resampling avoids holes but cannot invent occluded content. Keep movement small, inspect bending and reduce amplitude for difficult boundaries.

Does the local render spend generation funds?

No. It reads local media and renders locally. Missing media can be made through PopNow after an exact quote and approval; no paid call occurs in the renderer.