Omniverse Realtime Viewer

Canonical Stylized-Visualization Recipes

Use this contract when the user wants to recreate the named presentation styles, not merely produce related NPR looks. It defines visual behavior and numerical defaults without requiring a particular app layout or geometry authoring library.

All color values are normalized, display-referred RGB applied in SPG to native RGBA8 color unless a section says otherwise. The legacy BGRA equation ordering is a migration concern, not the graph output format. Line colors are premultiplied only during composition. The canonical reference resolution is 1920x1080 and the base line thickness is 1.35 px; at another resolution multiply widths and screen-space pattern sizes by viewport_height / 1080.

Contents

• Preset Manifest

• Shared Image Rules

• Image-Pass Defaults

• Guide-Driven Defaults

• Canonical Line Language

• Lighting Profile

Preset Manifest

This source catalog contains 14 presets. An explicit full-wireframe mode is an additional diagnostic mode, not a fifteenth source recipe or a renamed line-enable toggle. Declare its surface-fill and rear-edge policy separately.

Shared Image Rules

• SPG input is native RGBA8 LdrColor. The original presentation equations described BGRA transport bytes: port by named R/G/B channels without changing their math, and perform any BGRA swap only at the delivery boundary.

• Evaluate canonical image passes in float32 with top-left integer pixel coordinates. Convert each final RGB channel with uint8(floor(clamp(channel,0,1)*255 + 0.5)); never rewrite alpha.

• Luminance is 0.2126 R + 0.7152 G + 0.0722 B.

• Clamp after contrast, palette, and composite operations; preserve input alpha.

• Any neighborhood pass reads an immutable input and writes a separate output; use an internal intermediate where the original pass had a scratch boundary.

• A smooth gate uses t*t*(3-2*t) after clamping t to [0,1].

• Pixels with instance ID 0 are background for guide-driven presets.

• Screen patterns use integer pixel coordinates and deterministic phases.

Image-Pass Defaults

Studio And Solid

studio leaves OVRTX color unchanged. solid replaces visible mesh fill in the line/depth pass with RGB (0.43, 0.72, 0.78) at full opacity. Both retain the shared contour layer; neither injects post-process noise.

Toon

For each RGB channel, first apply c=clamp((c-0.5)*1.08+0.5). Compute luminance L, then saturation as c=L+(c-L)*1.14. Set band=floor(L*4+0.5)/4, lift=(band-L)*0.88, and add lift to every channel. Finally set temperature=(band-0.5)*0.10 and add (0.10,0.01,-0.08)*temperature to RGB before the shared clamp/quantize step.

Clay

Use clay RGB (0.76, 0.64, 0.49), contrast 1.08, cavity strength 1.35, and ridge strength 0.70. Compare center luminance to the four-neighbor mean. Darken valleys by valley * cavity_strength * 3.5, lift ridges by ridge * ridge_strength * 2.5, and clamp material gain to [0.12, 1.35]. Add a warm ridge highlight capped at 0.16, with green/blue contributions scaled by 0.72/0.38. More exactly, shade=clamp((L-0.5)*1.08+0.5), gain=clamp(0.43+0.82*shade-3.5*1.35*valley+2.5*0.70*ridge, 0.12,1.35), and highlight=min(0.16,ridge*0.70).

Preserve the environment with neutral_c=L+(c-L)*(1-0.88) and surface=smooth(clamp((L-0.20)/0.16,0,1)). The result is neutral + (clay_result-neutral)*surface. This is a presentation clay imitation, not a material-binding override.

Blueprint

Use six luminance levels and contrast 1.12. Map quantized tone to R=0.025+0.20*tone, G=0.090+0.57*tone, B=0.220+0.74*tone. Here L=clamp((luminance-0.5)*1.12+0.5) and tone=floor(L*5+0.5)/5. Draw a 24 px grid with strength 0.055; every fourth grid line is a major line at 1.8x strength. A line is exactly one pixel wide: set grid=0.055 when x mod 24 == 0 or y mod 24 == 0; overwrite it with 0.099 when x mod 96 == 0 or y mod 96 == 0. Intersections use the same overwritten value, not a sum. Add grid*(0.35,0.65,1.0) to RGB.

Halftone

Use 5 px square cells, contrast 1.08, saturation 1.16, and effect strength 0.88. Place one centered circular dot per cell with radius (1-luminance) * cell_size * 0.72. Paper RGB is (0.965, 0.935, 0.845); dot RGB is (0.035+0.58R, 0.025+0.50G, 0.045+0.52B). The canonical raw pass sets L=clamp((source_luminance-0.5)*1.08+0.5) and c=clamp(L+(source_c-L)*1.16) before computing the radius and dot color. It uses binary coverage: cell center is (2,2), and the dot is selected when (x mod 5-2)^2 + (y mod 5-2)^2 <= radius^2; there is no analytical antialiasing. Blend source toward selected paper/dot RGB by 0.88.

Hatch

Use spacing 8 px, line width 1 px, contrast 1.04, and strength 0.94. Filter tone with a five-sample cross at radius 2: center weight 0.40, four neighbors 0.15 each. Surface weight is 0.22 + 0.78 * clamp(chroma * 5). Add diagonal families at shadow thresholds 0.86 and 0.76 with feathers 0.075 and 0.10; add vertical/horizontal families at double spacing for thresholds 0.46/0.28 with feathers 0.09/0.08. Family strengths are 0.84, 0.90, 0.96, 1.0. Paper is warm near (0.91, 0.895, 0.845) with slight luminance lift; graphite starts at 0.075 + 0.08*luminance.

The four family masks are respectively (x+y) mod 8 < 1, (x-y+16384) mod 8 < 1, x mod 16 < 1, and y mod 16 < 1. Their gated ink values are the family strength times surface weight times smooth(clamp((threshold+feather-L)/(2*feather),0,1)); combine families with max, not addition. Set paper RGB to (0.91+0.075L+0.035(R-L), 0.895+0.070L+0.025(G-L), 0.845+0.065L+0.020(B-L)), set all graphite channels to 0.075+0.08L, interpolate paper toward graphite by ink, then blend source toward that result by 0.94.

Rim

First apply four softly transitioned tone bands with saturation 1.18, contrast 1.10, quantization 0.90, and cool-shadow strength 0.55. Derive the inner image-space rim from the filtered center versus the average of the three darker samples at radius 2. Use threshold 0.055, softness 0.10, strength 0.88, and rim RGB (0.72, 0.96, 1.0). When normal and ID guides are valid, add normal creases at strength 0.32 and ID boundaries at 0.72, toward RGB (0.68, 0.96, 1.0).

For the band pass, five-sample filtered luminance uses center weight 0.40 and four direct neighbors at 0.15. Set scaled=L*3, lower=floor(scaled), fraction=scaled-lower, transition=smooth((fraction-0.40)/0.20), and band=min(1,(lower+transition)/3). Add (band-center_luminance)*0.90 to saturated RGB, then add cool-shadow RGB (-0.05,+0.01,+0.08)*(1-band)*0.55. At radius 2, filter left, right, up, and down the same way; subtract the brightest of those four and divide the remaining sum by three. The rim signal is max(0,L-dark_surround) and rim coverage is min(1,smooth((signal-0.055)/0.10)*0.88).

Pixel

Use a 4x4 Bayer matrix with row-major ranks [0,8,2,10; 12,4,14,6; 3,11,1,9; 15,7,13,5]. Apply contrast 1.08, saturation 1.12, four levels per RGB channel, and full strength. This preset is ordered dither with a limited palette, not nearest-neighbor block enlargement. For each contrasted and saturated channel, set scaled=c*3, q=floor(scaled), and threshold=(rank+0.5)/16; increment q when scaled-q > threshold, clamp it to 3, and output q/3.

Guide-Driven Defaults

For all guide-driven styles, compute normal crease from the minimum dot product between the center view normal and its four direct neighbors. The canonical crease signal is a smooth gate of (1 - least_dot - 0.012) * 7.5. An ID boundary is 1 when a nonzero center ID differs from any direct neighbor.

AOV Ink

Use paper RGB (0.955, 0.935, 0.875). On surface pixels, blend a pigment wash with wash=0.50+0.48*luminance plus source RGB weights (0.20,0.18,0.16). Orient two mark fields with the absolute view-normal X/Y components, periods 8/9, mark widths 0.15/0.12, tone gates (0.82,0.18) and (0.58,0.16), and strengths 0.48/0.62. Fuse crease at 0.78 and ID boundary at 1.0. Ink RGB is (0.035, 0.075, 0.095). The exact fields are phase_a=(x*(0.72+abs(Nx))+y*(0.72+abs(Ny)))/8 and phase_b=(x*(0.72+abs(Ny))-y*(0.72+abs(Nx)))/9. Mark A exists when fract(phase_a)<0.15 and has gate smooth((0.82-L)/0.18); mark B uses fract(phase_b)<0.12 and smooth((0.58-L)/0.16). Combine both marks, crease, and boundary with max, multiply by the nonzero-ID surface mask, then interpolate the pigment wash toward ink by that value.

Facets

Compute orientation as clamp(0.5 + 0.5*(0.42Nx - 0.48Ny + 0.76Nz)), then form four softly transitioned bands by setting scaled=orientation*4, lower=floor(scaled), fraction=scaled-lower, blend=smooth((fraction-0.36)/0.28), and band=min(1,(lower+blend)/4). Base RGB is (0.10+0.70b+0.20max(Nx,0), 0.16+0.52b+0.14max(Ny,0), 0.30+0.34b+0.22max(-Nx,0)), plus 0.08 of source RGB and gain 0.86+0.18*luminance. Fuse crease at 0.72 and full ID boundaries toward RGB (0.025,0.035,0.070). This is a view-normal plane treatment; it does not claim to reconstruct geometric face IDs. For background, use neutral=0.94+(luminance-0.5)*0.10 and replace all three channels with that neutral value.

Technical

Use form clamp(0.58 + 0.30Nz - 0.12Ny) and stock RGB (0.68+0.18f, 0.78+0.14f, 0.84+0.12f). Fuse normal crease at 0.82 and ID boundary at 0.92 toward RGB (0.025,0.055,0.085). Preserve nonsurface pixels as neutral=clamp(0.955+(luminance-0.5)*0.07,0.90,0.985).

Concept

Use smooth form 0.5 + 0.5*(0.38Nx - 0.46Ny + 0.80Nz). Interpolate from cool RGB (0.18,0.42,0.72) to warm RGB (0.96,0.69,0.27), then add 0.06 of source RGB. Fuse crease at 0.74 and ID boundary at 0.92 toward (0.035,0.045,0.075).

X-Ray

This is presentation X-ray. Use rim smooth((1-abs(Nz)-0.08)/0.72) and core 0.16+0.17*max(Nz,0). On surfaces, move neutral background toward RGB (0.54+core+0.12rim, 0.82+0.35core+0.10rim, 0.90+0.20core+0.08rim). Fuse crease at 0.58 and ID boundary at 0.92 toward RGB (0.015,0.30,0.42). Rear topology comes only from the hidden-line pass.

Keep these values as the canonical preset recipe. If the look targets ordinary tessellated USD meshes and must retain stronger surface form while showing support/object lines through occluders, use the separate optional mesh-xray.md delivery profile. It does not alter this recipe or any BREP-aware path.

Canonical Line Language

feature width = 1.35 * width_scale; silhouette width = feature width * silhouette_scale. Pattern periods are pixels at 1080p. Silhouettes are always solid.

Technical hidden lines are dotted with width scale 0.58, period 8.5, and RGBA (0.30,0.40,0.48,0.46). X-Ray hidden lines are dashed with width scale 0.62, period 11.0, duty 0.48, and RGBA (0.08,0.52,0.62,0.52).

Use these visible feature/silhouette RGBA pairs respectively:

Lighting Profile

The canonical viewer-owned studio rig uses three distant lights:

Author this only in a viewer-owned session/composite scope. Keep existing stage lighting as the interactive-app default unless the user enables the studio profile. Preset selection never changes lighting implicitly. Canonical reproduction capture is the exception: it starts from the neutral fixture, enables only this rig for all 14 presets, records lighting_profile=studio, and removes the temporary rig afterward.