ANR Tools

ANR Depth Camera

Moves the camera a little after the shot, with real parallax from the depth: small pushes, pans, dolly zooms, or a second stereo eye.

  • Utility
  • Comp & Depth pack
  • 2 inputs
  • 17 controls

Overview

Depth Camera re-photographs the plate from a new camera position. Every pixel is placed in 3D at its distance from ANR Depth Prep and seen again from the camera you set, so near things shift more than far ones, as with a real move. Use it for a slight push-in or drift, a reframe with parallax, a dolly zoom, or a left and right eye for stereo.

A moved camera sees a little of what the original never did, behind the edges of foreground objects. The node fills those gaps from the farther side, so the background stretches in and the foreground never smears; Output > Hole Matte shows exactly where, for a paint or inpaint fix.

Output > ST Map gives the warp itself, so Nuke's STMap node can apply the same move to mattes or other passes.

Inputs

SourceDepthDepth CameraOutput
  1. Sourcerequired

    The plate to move. All four channels move with it.

  2. Depthrequired

    An ANR Depth Prep with Output on Depth; its lens and Focus Distance are read through the wire. Any other image is read as distance from its red channel, with the Custom lens.

Output. The plate seen from the new camera (or the chosen eye). Output can show the Hole Matte (white where the plate had nothing to show) or the ST Map of the warp (red and green: where to read the plate; alpha: not a hole).

Quick start

  1. Make a distance pass with ANR Depth Prep (Output on Depth) and set its lens in Lens if you know it.
  2. Add ANR Depth Camera. Connect the plate to Source and the Depth Prep to Depth.
  3. Open Camera Move and animate Translate (negative Z pushes in) or Rotate for the move. Keep it small next to the nearest distance in the shot.
  4. Set Output to Hole Matte to see what the move uncovered, then back to Result.
  5. Paint or patch the holes if they show, or set Gaps to Leave Empty and fill them your own way.

The panel

The node's panel as Nuke shows it, with the values of a new node. Click any control to see what it does; tabs and sections work as in Nuke.

ANR_Depth_Camera1
StatusConnect an ANR Depth Prep (Output = Depth) to the Depth input.
ModeCamera Move
OutputResult
GapsStretch Background
Edge Cleanup0.08
Translatex0y0z0
Rotatex0y0z0
Zoom1
Dolly Zoom0
Keep Size At0
EyeRight
The Plate IsBetween The Eyes
Interaxial0.065
Convergence0
LensFrom Depth Prep
Focal Length35
Horizontal Aperture36

All controls

ANR Depth Camera17 controls
StatusStatusstatus

Where the lens comes from, and what to expect.

ModeMenudcMode

Camera Move: a new camera position and angle. Stereo: a left / right eye.

Default: Camera MoveOptions: Camera Move Stereo

OutputMenudcOutput

Result: the plate from the new camera.
Hole Matte: white where the new camera sees something the plate never did -- paint or inpaint there.
ST Map: the warp itself (red, green = where to read the plate; alpha = not a hole), for Nuke's STMap node on other passes.

Default: ResultOptions: Result Hole Matte ST Map

GapsMenudcFill

Stretch Background: fill uncovered gaps from the farther side. Leave Empty: black, alpha 0, for your own fill.

Default: Stretch BackgroundOptions: Stretch Background Leave Empty

Edge CleanupSliderdcEdgeCleanup

Drop the soft, mixed pixels on a depth edge (a jump bigger than this fraction) so they do not drag a halo across the gap. 0 keeps them.

Default: 0.08

Camera Move

TranslateXYZdcTranslate

Move the camera, in the depth's units: X right, Y up, Z back (negative Z pushes in). Animate it for a move.

Default: 0 0 0

RotateXYZdcRotate

Tilt (X), pan (Y) and roll (Z) in degrees, about the camera itself.

Default: 0 0 0

ZoomSliderdcZoom

Multiply the focal length.

Default: 1

Dolly ZoomSliderdcDollyZoom

Push in by this much while widening the lens so that what stands at Keep Size At stays the same size -- the background stretches away. Negative pulls out.

Default: 0

Keep Size AtSliderdcDollyKeep

The distance that keeps its size in a Dolly Zoom. 0 uses the Depth Prep's Focus Distance, else the middle of the frame's depth.

Default: 0

Stereo

EyeMenudcEye

Which eye to render. Anaglyph (red / cyan) and Side by Side are for checking.

Default: RightOptions: Left Right Anaglyph Side by Side

The Plate IsMenudcStereoBase

Between The Eyes: both eyes are made, each half as far off (smaller gaps). The Left Eye: the left eye is the plate untouched, only the right is made.

Default: Between The EyesOptions: Between The Eyes The Left Eye

InteraxialSliderdcInteraxial

Distance between the eyes, in the depth's units (0.065 m is human).

Default: 0.065

ConvergenceSliderdcConvergence

The distance that sits at the screen. 0 uses the Depth Prep's Focus Distance, else the middle of the frame's depth.

Default: 0

Lens

LensMenudcLensFrom

From Depth Prep: use the Focal Length and Aperture on the Depth Prep feeding the Depth input.

Default: From Depth PrepOptions: From Depth Prep Custom

Focal LengthSliderdcFocal

Custom lens, mm.

Default: 35

Horizontal ApertureSliderdcAperture

Custom film back, mm.

Default: 36

Tips

  • For a dolly zoom, set Dolly Zoom and Keep Size At to the subject's distance: the subject keeps its size while the background stretches away. With Keep Size At at 0, the Depth Prep's Focus Distance is used.
  • For stereo, set Mode to Stereo. Eye > Anaglyph or Side by Side shows both eyes for checking; render Left and Right for delivery.
  • The Plate Is > Between The Eyes makes both eyes, each half as far off, which keeps the gaps smaller. The Left Eye keeps the plate as the left eye and makes only the right.
  • Convergence is the distance that sits on the screen; at 0 the Depth Prep's Focus Distance is used.
  • If a halo of edge pixels drags across a gap, raise Edge Cleanup; at 0 the soft edge pixels are kept.
  • Calibrate the Depth Prep in metres so that Translate and Interaxial are in metres too (0.065 is a human eye distance).

Good to know

  • The node runs no AI model and needs no graphics card.
  • The plate only holds what the original camera saw: big moves open big gaps, and the fill is a stretch of the background, not new picture.
  • In Camera Move with no move set (Translate, Rotate and Dolly Zoom at 0, Zoom at 1) and Output on Result, the plate passes through unchanged.

Something unclear or missing? Write to [email protected].