Node-based multi-pass compositing
Rebuild a beauty from CG passes and integrate elements in node graphs.
20 project-authored proposals · taxonomy v1.1 · 0 new benchmark runs. Verifiers and partial-credit rules below are designs, not executed results.
- Typical applications
- Nuke · Fusion · Blender compositor
- Class GUI-need label
HY· Hybrid
Task proposals · 20
Expand to inspect the complete original wording and proposed checks. Original proposal text remains English in both interfaces.
R06.05.02.05-T01 Rebuild the beauty from light AOVs in Nuke
Original instruction
Open shots/sh010/sh010_comp_v001.nk. Read fixtures/sh010_cg_v003.####.exr (frames 1001-1024, multichannel: diffuse_direct/indirect, specular_direct/indirect, sss, transmission, emission, beauty). Build a Shuffle + Merge(plus) tree that rebuilds the beauty from the 7 light AOVs only, and render it with a Write node to render/sh010_rebuild.####.exr (16-bit half, ZIP1). Save as sh010_comp_v002.nk.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Per frame RMS(rebuild - beauty) <= 1e-3 on RGB. The .nk is parsed: the beauty channel must not feed the Write (graph walk from Write), all 7 AOVs reach a plus-merge. Write node knobs checked for file type, datatype and compression. Frame range 1001-1024 complete.
Proposed partial credit
30% all 7 AOVs shuffled out; 40% rebuild RMS <= 1e-3 on all frames; 15% beauty not used upstream; 15% Write spec and frame range exact.
R06.05.02.05-T02 Boost car specular via Cryptomatte only
Original instruction
In fixtures/sh020_comp.nk the client wants the car paint to read glossier. Using the Cryptomatte layer in sh020_cg.####.exr, raise specular_direct + specular_indirect by exactly 25% on objects matching 'car_body*' only, leave every other pixel unchanged, and render render/sh020_v002.####.exr (frames 1-48).
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clickdrag· Drag & dropvis· Visual judgment- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Inside the GT car mask (coverage > 0.99) output = beauty + 0.25*spec within 1e-3; outside the mask (coverage < 0.01) RMS vs beauty <= 1e-4; matte-edge band checked for no dark halo (alpha-weighted error <= 2e-3). .nk must contain a Cryptomatte node whose matteList names car_body objects.
Proposed partial credit
20% Cryptomatte node with correct matte list; 50% spec gain 1.25 inside mask; 30% outside pixels and edges untouched.
R06.05.02.05-T03 Add Z-depth defocus in the Blender compositor
Original instruction
Open fixtures/interior_shot.blend (Cycles, Z pass enabled, already rendered to the Render Result cache). In the compositor add a Defocus driven by the Depth pass so the plane at 3.2 m is in focus, max blur 12 px, f-stop 2.8 equivalent. Save the file and render frame 1 to render/interior_dof.exr.
- Platforms
mac· macOSwin· Windows- Applications
- Blender
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Saved .blend compositor tree inspected: Defocus node fed by the Depth socket, max blur 12, focus distance 3.2 m (via camera DOF or node). Render compared with a reference made with the same parameters: SSIM >= 0.97; in the GT in-focus slab (3.1-3.3 m) Laplacian variance within 5% of the sharp input.
Proposed partial credit
30% Defocus node wired to Depth; 30% parameters exact; 40% render matches reference SSIM >= 0.97.
R06.05.02.05-T04 Integrate a CG car onto a street plate with shadow catcher
Original instruction
Comp fixtures/sh030_car_cg.####.exr (beauty, shadow_catcher, alpha) over plate fixtures/sh030_plate.####.dpx (frames 1-72, ARRI LogC3). Linearize the plate, comp the premultiplied car, multiply the shadow pass into the plate, match CG black level to the plate, and write render/sh030_comp.####.exr. No dark edge fringe allowed.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke · Fusion
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentclick· Point & click- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
Output compared with a supervisor reference: SSIM >= 0.95 per frame. Edge band (alpha 0.05-0.95) luminance ratio to reference within 3% (no fringe). Shadow region (GT mask) darkening ratio within 5% of reference. Car 1st-percentile luma within 2% of plate's. Read node colorspace must be LogC3.
Proposed partial credit
20% plate linearized correctly; 30% car premult comp edge-clean; 25% shadow integrated; 25% black-level match.
R06.05.02.05-T05 Relight a CG head with a rim light from N and P passes
Original instruction
The director wants a cool rim light from screen-left on the CG head in fixtures/sh040_head.####.exr (has N, P, albedo, beauty AOVs). Using a Relight setup (Nuke Relight or normal-pass math), add a rim of color #9DB8FF so the GT rim mask brightens by 0.8 stops on average, with no change where N faces the camera. Render render/sh040_rim.####.exr (frames 1-30).
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropvis· Visual judgmentgeom· Precise geometry- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
In the GT rim mask (N.x < -0.6) mean luminance gain = 0.8 +/- 0.1 stops and hue of the added light within 5 deg of #9DB8FF. Facing-camera region (N.z > 0.9) RMS change <= 2e-3. Light direction recovered from the output by regression on N within 15 deg of screen-left. Per-frame flicker std <= 0.02 stops.
Proposed partial credit
25% relight setup reads N/P; 35% rim strength and hue in band; 25% front-facing untouched; 15% temporally stable.
R06.05.02.05-T06 Rebalance Cycles light groups to a 3:1 key-to-fill ratio
Original instruction
fixtures/portrait_lightgroups.blend renders light groups key, fill, rim and env. In the Blender compositor, scale the groups so the GT face mask has key:fill luminance ratio 3.0 and total face exposure equals ref/face_target.exr within 0.05 stops; rim and env stay at gain 1. Save and render render/portrait_relit.exr.
- Platforms
mac· macOSwin· Windows- Applications
- Blender
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Compositor tree parsed: Mix/Math nodes scale each light-group pass and sum them; rim/env gains = 1.0. On the render, key:fill ratio in the face mask = 3.0 +/- 0.15 (computed from per-group gains x pass means); face exposure within 0.05 stops of target. Non-light-group pixels unchanged.
Proposed partial credit
30% light groups summed correctly; 40% ratio 3:1 met; 30% exposure target met.
R06.05.02.05-T07 Diagnose and fix double premultiplication halos
Original instruction
fixtures/sh050_comp_broken.nk shows dark halos around a CG drone over a sky plate. Find every cause in the graph (expected: a Grade applied after Premult and a second Premult downstream) and fix it without re-rendering CG. Save sh050_comp_fixed.nk and render frame 1100 to render/sh050_1100.exr.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
vis· Visual judgmentclick· Point & clickdrag· Drag & drop- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro apphidden· Hidden/stale state- Proposed difficulty
- 3 / 5
Proposed verifier
Rendered frame vs GT clean comp: RMS <= 1e-3; edge-band (alpha 0.05-0.95) mean luminance deficit <= 0.5%. Graph walk on the saved .nk: exactly one Premult on the CG branch, grades happen on unpremultiplied data (Unpremult->Grade->Premult or Grade (unpremult by alpha)). No Read node changed. Nuke's Script Editor and terminal mode (nuke -t) are disabled, and .nk writes from outside Nuke are logged and fail.
Proposed partial credit
40% halo removed in render; 40% graph premult order correct; 20% no other nodes altered.
R06.05.02.05-T08 Set up ACES color management for a mixed-source comp
Original instruction
In a new Nuke script using the bundled ACES 1.3 studio OCIO config, read: fixtures/cg_acescg.####.exr (ACEScg), fixtures/plate_logc4.####.dpx (ARRI LogC4/AWG4) and fixtures/logo_srgb.png (sRGB texture). Comp the logo and CG over the plate, then write render/master.####.exr in ACES2065-1 and review/master_rec709.mov (ProRes 422, Rec.709 display).
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsvis· Visual judgment- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Read/Write colorspace knobs and OCIO config path parsed from the .nk. A Macbeth chart frame embedded in each source is measured in the EXR output: ΔE2000 <= 0.5 vs the reference transform chain. MOV probed with ffprobe for codec, primaries tag bt709 and chart ΔE2000 <= 1.5.
Proposed partial credit
25% OCIO config set; 35% each Read colorspace correct; 25% EXR AP0 values match; 15% Rec.709 review encode correct.
R06.05.02.05-T09 Deep-composite a creature through volumetric smoke
Original instruction
Comp the creature deep render fixtures/sh060_creature_deep.####.exr into deep smoke fixtures/sh060_smoke_deep.####.exr (frames 1-40) using DeepMerge and DeepHoldout so the creature is correctly partially occluded by smoke at every depth, then flatten with DeepToImage over plate sh060_plate.####.exr. Write render/sh060_comp.####.exr.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Per-frame RMS vs GT deep-flattened comp <= 2e-3; in the GT interpenetration region (creature inside smoke, depth overlap) error <= 4e-3 and no hard edge (gradient energy <= 1.2x GT). .nk graph must contain DeepMerge/DeepHoldout and DeepToImage; flattening the two layers as 2D then merging fails the overlap check.
Proposed partial credit
25% deep reads and DeepMerge correct; 35% holdout depth-correct in overlap region; 25% flattened over plate; 15% full frame range.
R06.05.02.05-T10 Add a glow to the emission pass in Fusion
Original instruction
In DaVinci Resolve's Fusion page, open the comp for clip neon_sign.exr (has beauty and emission AOVs). Extract the emission AOV, apply a Glow (size 20, gain 0.5, Gaussian), and add it back over the beauty with a Merge set to Add. Render the clip from the Deliver page to render/neon_glow.exr.
- Platforms
mac· macOSwin· Windows- Applications
- DaVinci Resolve (Fusion)
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & click- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 2 / 5
Proposed verifier
Exported .comp (Lua table) parsed: Glow tool on the emission channel with XGlowSize 20, Gain 0.5, Merge apply mode Add. Output EXR vs reference: RMS <= 2e-3. Non-emissive GT region (emission == 0 and > 40 px from emitters) unchanged within 1e-4.
Proposed partial credit
30% emission isolated; 30% glow parameters exact; 40% render matches reference.
R06.05.02.05-T11 Denoise passes separately and rebuild the beauty
Original instruction
fixtures/kitchen_64spp.blend has a 64-sample Cycles render with Denoising Data, diffuse and glossy direct/indirect/color passes. In the compositor, denoise the diffuse and glossy light passes separately (Denoise node with albedo and normal), rebuild the beauty, and render render/kitchen_denoised.exr. Do not raise the sample count.
- Platforms
mac· macOSwin· Windows- Applications
- Blender
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
- None declared
- Proposed difficulty
- 3 / 5
Proposed verifier
Scene samples must still be 64 (checked in saved .blend). Output vs 4096-spp GT: PSNR >= 38 dB and higher than single-node beauty denoise baseline by >= 0.5 dB. Compositor graph must show separate Denoise nodes on diffuse and glossy branches combined by Multiply (color) and Add.
Proposed partial credit
20% sample count unchanged; 30% passes denoised separately; 30% beauty rebuilt correctly; 20% PSNR targets.
R06.05.02.05-T12 Package an AOV rebuild gizmo and apply it to 3 shots
Original instruction
Build a Nuke gizmo 'AOV_Rebuild' that rebuilds beauty from 7 light AOVs and exposes float knobs diffuse_gain, spec_gain, sss_gain, emission_gain (default 1). Save it to tools/AOV_Rebuild.gizmo, then use it in sh070, sh080, sh090 (fixtures/*.nk) and render each with spec_gain 1.2 as listed in notes/client_notes.txt.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clicktype· Text & forms- Work product
WP06· Code/engineering- Conditions
licensed· Licensed pro appmultiapp· Multi-app- Proposed difficulty
- 4 / 5
Proposed verifier
Gizmo file parsed for the 4 knobs with defaults 1.0. Each shot render with gain defaults (re-rendered by the grader) equals its beauty RMS <= 1e-3. Delivered renders match spec_gain 1.2 reference RMS <= 1e-3. Each .nk uses the gizmo (class AOV_Rebuild), not copied raw nodes.
Proposed partial credit
30% gizmo with exposed knobs; 30% identity at defaults; 30% three shots rendered with gain 1.2; 10% gizmo instanced in each script.
R06.05.02.05-T13 Add vector motion blur from the motion pass
Original instruction
The CG wheel render fixtures/wheel_nomb.####.exr (frames 1-24) has no motion blur but includes a motion (velocity) pass. Add 2D motion blur equivalent to a 180-degree shutter with VectorBlur (Nuke) or Vector Blur (Blender), comp over the plate, and render render/wheel_mb.####.exr.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke · Blender
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
drag· Drag & dropclick· Point & clickvis· Visual judgment- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 3 / 5
Proposed verifier
Compared to a GT 3D-motion-blurred render: SSIM >= 0.93 inside the GT moving mask, static regions RMS <= 1e-3. Blur length on the rim measured along tangent within 15% of GT. Project parsed for the vector-blur node with motion channel connected and shutter/scale equivalent to 0.5.
Proposed partial credit
30% motion pass wired; 30% shutter scale correct; 40% render matches 3D blur GT.
R06.05.02.05-T14 Add depth fog sampled from the plate sky
Original instruction
Comp the CG castle fixtures/sh100_castle.####.exr (with Z) over plate sh100_plate.####.exr. Add atmospheric fog using the Z pass so objects at 400 m reach 60% fog opacity, with fog color sampled from the plate's horizon haze (use the color picker on the haze band). Render render/sh100_comp.####.exr.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke · Fusion
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
vis· Visual judgmentclick· Point & clickdrag· Drag & drop- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
Fog color in output (estimated from far CG pixels by regression) within ΔE2000 <= 3 of GT haze sample. Fog opacity vs depth fit: 60% +/- 5% at 400 m, monotonic, < 5% at 20 m. Plate pixels outside CG alpha unchanged (RMS <= 1e-4). VLM rubric only for overall integration (10%). Nuke Script Editor/terminal mode and Fusion Console/fuscript are disabled; renders or comps written outside the app are logged and fail.
Proposed partial credit
30% fog driven by Z; 30% opacity curve; 30% haze color match; 10% integration rubric.
R06.05.02.05-T15 Roto a pillar to hold out a CG creature
Original instruction
The CG creature in fixtures/sh110_creature.####.exr walks behind a real pillar in plate sh110_plate.####.exr (frames 1-48), but the render has no holdout. Draw and animate a Roto spline for the pillar (tracking allowed), use it to hold out the creature, and render render/sh110_comp.####.exr plus the matte render/sh110_pillar_matte.####.png.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke · Fusion
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
path· Continuous pathgeom· Precise geometryvis· Visual judgment- Work product
WP10· Annotations- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 3 / 5
Proposed verifier
Pillar matte IoU >= 0.93 vs GT on every frame and boundary F-score (2 px) >= 0.85. Comp RMS vs GT comp <= 3e-3 in the overlap region. Saved script must contain a Roto/Polygon spline with >= 6 keyframes; imported mattes or AI segmentation nodes not present in the starting toolset are flagged. Nuke Script Editor and Fusion Console are disabled; .nk/.comp edits outside the app are logged.
Proposed partial credit
40% matte IoU threshold; 20% boundary accuracy; 30% holdout comp correct; 10% roto keyed in-app.
R06.05.02.05-T16 Undistort, comp and redistort with matched grain
Original instruction
Plate fixtures/sh120_plate.####.dpx was shot on an anamorphic lens; a grid chart fixtures/lens_grid.dpx is provided. Solve lens distortion from the grid, undistort the plate, comp the undistorted CG sh120_cg.####.exr, redistort the result, regrain to match plate grain per channel, and write render/sh120_final.####.exr.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
vis· Visual judgmentclick· Point & clickdrag· Drag & dropgeom· Precise geometry- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro applong· Long-horizon- Proposed difficulty
- 5 / 5
Proposed verifier
Grid detection on the redistorted grid round-trip: corner reprojection RMS <= 0.5 px. CG feature points (GT) land within 1 px of GT redistorted positions. Grain: per-channel noise std in 6 flat CG patches within 10% of plate grain and spectral slope within 15%. Plate pixels outside CG unchanged after round-trip (PSNR >= 45 dB).
Proposed partial credit
25% distortion solve accuracy; 25% CG placed in distorted space; 30% grain match; 20% plate fidelity.
R06.05.02.05-T17 Re-version a comp to renamed AOVs in a new render
Original instruction
Lighting delivered sh130_cg_v005.####.exr, which renames AOVs (diffuse_direct->diff_dir, specular_indirect->spec_ind etc., see notes/aov_rename.csv) and adds a coat AOV. Update fixtures/sh130_comp_v004.nk so every Read points at v005, every broken Shuffle uses the new names, coat is added to the rebuild, and render render/sh130_v005.####.exr.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
click· Point & clicktype· Text & formsvis· Visual judgment- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro apphidden· Hidden/stale state- Proposed difficulty
- 3 / 5
Proposed verifier
Saved .nk parsed: all Reads reference v005 paths, no Shuffle references a missing layer (grader opens the script in nuke -t and checks for errors), coat included in the plus tree. Output vs v005 beauty RMS <= 1e-3 on all frames. Other grading nodes and their knob values unchanged from v004.
Proposed partial credit
25% Reads re-pathed; 30% shuffles repaired; 20% coat added; 25% render equals v005 beauty and grades preserved.
R06.05.02.05-T18 Headless AOV rebuild via Nuke Python (control pair of T01)
Original instruction
Control pair of T01: from a terminal only, write tools/rebuild_sh010.py that builds the same 7-AOV plus tree for fixtures/sh010_cg_v003.####.exr and renders render/sh010_rebuild_cli.####.exr (frames 1001-1024, 16-bit half, ZIP1) with nuke -t. No GUI session may be opened.
- Platforms
mac· macOSwin· Windowslin· Linux- Applications
- Nuke (nuke -t)
- Interfaces
term· Terminal- GUI need · task setting
CL· CLI-sufficient- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
code· Code & commands- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 2 / 5
Proposed verifier
Same thresholds as T01: per-frame RMS(rebuild - beauty) <= 1e-3, beauty not upstream, Write spec exact. The license server log must show only terminal (nuke -t) checkouts during the attempt; GUI checkouts invalidate the run. Scored side by side with T01 to measure the GUI-vs-CLI delta.
Proposed partial credit
30% script runs without error; 40% rebuild RMS <= 1e-3; 30% Write spec and frame range exact.
R06.05.02.05-T19 Grade three products to client swatches via Cryptomatte
Original instruction
In fixtures/packshot.comp (Fusion Studio) the bottle, cap and label (Cryptomatte IDs) must match the swatches in input/client_swatches.json under the shot's lighting. Isolate each with a Cryptomatte tool, color-correct so the mean lit color in each mask matches its swatch, and render render/packshot_graded.exr.
- Platforms
mac· macOSwin· Windows- Applications
- Fusion Studio · Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
G2· GUI-essential- GUI need · ordinary access
CL· CLI-sufficient- Interaction primitives
vis· Visual judgmentclick· Point & clickdrag· Drag & drop- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro app- Proposed difficulty
- 4 / 5
Proposed verifier
In each GT object mask, mean Lab color within ΔE2000 <= 2 of the swatch target, and luminance texture preserved (correlation of high-pass detail >= 0.95). Background RMS change <= 1e-4. .comp parsed for 3 Cryptomatte tools and corrector nodes; no paint or bitmap replacement tools on the branches. Fusion Console/fuscript and Nuke Script Editor are disabled; .comp/.nk writes outside the app are logged and fail.
Proposed partial credit
15% three mattes isolated; 60% mean color within ΔE2000 <= 2 (20 points per object); 25% detail and background preserved.
R06.05.02.05-T20 Integrate a CG spacecraft into an aerial plate end to end
Original instruction
Supervisor brief notes/sh140_brief.pdf: comp fixtures/sh140_ship.####.exr (AOVs, Cryptomatte, Z, motion) over aerial plate sh140_plate.####.dpx (frames 1-96). Rebuild from AOVs, match black/white points and chroma to the plate using the grey-ball pass, add depth haze, vector motion blur and matched grain. Deliver render/sh140_final.####.exr and review/sh140.mov.
- Platforms
mac· macOSwin· Windows- Applications
- Nuke
- Interfaces
desk· Desktop GUI- GUI need · task setting
HY· Hybrid- GUI need · ordinary access
G2· GUI-essential- Interaction primitives
drag· Drag & dropvis· Visual judgmentclick· Point & click- Work product
WP05· Time-based media- Conditions
licensed· Licensed pro applong· Long-horizonmultiapp· Multi-app- Proposed difficulty
- 5 / 5
Proposed verifier
Deterministic: grey-ball pass in comp neutral within ΔE2000 <= 2 of plate grey card; ship 1st/99th luma percentiles within 3% of plate's; haze opacity at GT depth within 5%; grain std within 10%; motion blur SSIM >= 0.9 vs GT. VLM rubric (calibrated on supervisor ratings) for overall integration, 15% weight.
Proposed partial credit
25% AOV rebuild and plate match; 25% haze and motion blur; 20% grain; 30% deliverables spec (15 points) and integration rubric (15 points).
Mapped research references
No benchmark reference is assigned to this class in the current mapping.