Skip to main content
Card 3D

Card 3D · screenshot coming soon

​
What this node does

This node maps a COMFYTV_VIDEO clip onto a flat plane and transforms that plane in 3D — translating, rotating, and scaling it under a perspective camera — then renders the result back to 2D. Because it’s a true 3D projection (not a flat 2D warp), tilting the card produces correct perspective foreshortening, like the Card node in Nuke or a 3D layer in After Effects. It runs on a Torch backend as an fx-spec pass. If there’s no translation, no rotation, and scale is 1.0, it passes the clip through untouched. Input and output are both COMFYTV_VIDEO. To hand the result to native ComfyUI nodes, insert a Bridge — see the bridge note.

​
When to use it

  • Fake a 3D camera move — swing footage in perspective for a title card or reveal.
  • Tilt a plate to sit on a virtual floor/wall in a composite.
  • Build a simple parallax or “photo in space” look.

​
Parameters

​
fov

Camera field of view in degrees, 5 to 140 (default 40). Lower is a longer/telephoto lens (flatter perspective); higher is wide-angle (stronger perspective as the card tilts).

​
tx / ty / tz

Translate the card in 3D. tx and ty move it horizontally/vertically, each -4.0 to 4.0 (default 0). tz moves it toward or away from the camera, -0.95 to 6.0 (default 0) — negative pulls it closer (bigger), positive pushes it back (smaller).

​
rx / ry / rz

Rotate the card in degrees. rx (pitch) and ry (yaw) each -89 to 89 (default 0) tilt the card in perspective; rz (roll) -180 to 180 (default 0) spins it in-plane.

​
card_scale

Overall scale of the card, 0.05 to 8.0 (default 1.0). Resize the plane before it’s placed in the scene.

​
Outputs

​
Tips

  • Keep rx/ry within ±89 — the card can’t rotate to fully edge-on.
  • Use a lower fov (longer lens) when you want subtle perspective, higher fov for a dramatic tilt.
  • Combine tz with rotation to sell a card swinging in from depth.
  • Z Defocus — add depth-of-field so a tilted card reads even more 3D.
  • 360 Projection — for spherical rather than flat-plane 3D transforms.