{ "title": "What makes this landscape look photographic", "date": "2026-11-06", "status": "Research-backed implementation audit; photographic parity and sustained 1080p60 are not achieved", "overview": "We are building a continuous desert weather observatory: a coherent sandstone basin, clouds that form and dissipate, changing sunlight, and a camera that can travel from the ground through the weather to above it. The current renderer has substantial physical machinery, but its landforms, materials, coupled illumination and temporal image formation do not yet form a convincing photographic whole. More effects or higher resolution alone will not close that gap.", "scope": "A broad coverage map for this landscape and its planned day/twilight timelapse, not a claim to enumerate every optical phenomenon in nature. Each essential layer needs an implemented and validated solution; weather- or lens-dependent effects are enabled only when they belong in the shot.", "files": [ "terrain-material-lighting-audit.json", "atmosphere-cloud-lighting-matrix.json", "camera-temporal-display-matrix.json" ], "currentImage": "/outputs/observatory/v4-height-r32-float32-views1/ground-reference.png", "directionImage": "/observatory/assets/lookdev/ground-reference.png", "directionImageWarning": "Generated visual direction. This is not a real photograph, Sunburst output, or calibrated physical ground truth.", "annotatedCurrent": [ { "id": "01", "label": "Cliff geology", "rect": [ 0.169, 0.487, 0.48, 0.168 ], "detail": "Broad fluted walls and smooth slopes lack convincing stratified masses, fractures, ledges and erosion-linked debris. A heightfield or a rounded scan enlarged into a cliff cannot supply the needed geology by texture alone." }, { "id": "02", "label": "Foreground seam", "rect": [ 0.416, 0.783, 0.018, 0.217 ], "detail": "This research baseline predates the gravel-normal correction. Fixed-origin yaw, sharing-off and forward controls locate the signal in world-anchored debris. CPU ray queries identified separate stones with incorrect smoothed normals; corrected native captures improve their lighting. The full perceived line is not yet declared resolved." }, { "id": "03", "label": "Surface organization", "rect": [ 0.021, 0.602, 0.46, 0.154 ], "detail": "Large smooth areas and broadly repeated material character need a coherent hierarchy of sediment, rock fragments, runoff and weathering. Texture density and feature size must match the landform and camera distance." }, { "id": "04", "label": "Cloud form and lighting", "rect": [ 0.359, 0.032, 0.366, 0.476 ], "detail": "The hero cloud has large bulbous forms, a broad shelf and a strong crown-to-base transition. Native voxel samples already exist. Morphology, water-droplet phase behavior, environmental lighting and live reconstruction each need separate comparisons." }, { "id": "05", "label": "Depth and coupled light", "rect": [ 0.467, 0.542, 0.215, 0.112 ], "detail": "Distance, terrain shadow, cloud shadow, sky illumination and air scattering must agree. The image suggests limited depth and material richness, but a global fog or ambient boost would conceal rather than diagnose the underlying mismatch." } ], "pipeline": [ { "stage": "1. Physical scene", "items": [ "Geological macro form", "Erosion and stratigraphy", "Rock and sediment scale", "Correlated material maps", "Cloud morphology and weather" ] }, { "stage": "2. Incident light", "items": [ "Sun spectrum, disk and position", "Rayleigh, aerosol and ozone atmosphere", "Multiple-scattered sky", "Cloud and terrain visibility", "Sky and ground illumination of clouds" ] }, { "stage": "3. Light transport", "items": [ "Surface BRDF and normal filtering", "Contact and soft shadows", "Directional diffuse and specular environment", "Ground interreflection", "Cloud phase function and multiple scattering", "Air/cloud/terrain composition" ] }, { "stage": "4. Camera", "items": [ "Projection and physical scale", "Exposure and white balance", "Finite shutter", "Focus and lens response when visible", "Scene-linear color transforms" ] }, { "stage": "5. Reconstruction and delivery", "items": [ "Texture and shading footprints", "Temporal history and motion", "Stable pacing", "Tone/gamut mapping", "Display encoding", "Film cadence and compression" ] } ], "mostImportantFindings": [ "Numerically accurate evaluation of an approximate physical model does not establish that the model looks like real water clouds or real sandstone.", "The foreground line investigation found incorrect microgravel normals: averaging upper and lower faces turns some sunlit facets away from the Sun. A position-preserving faceted-normal correction improves native captures. Residual visual alignment near the slab, full motion and performance remain open.", "Cloud sunset flashes are traced to sparse external solar optical-depth sampling. A denser oracle is smooth; the production repair is not yet installed.", "The existing cloud atlas preserves native samples according to its source manifest. Source voxel resolution, incident-light cache resolution, ray integration and temporal reconstruction are different limits.", "Terrain shape and material organization are major visual limits. More triangles and larger textures do not create stratigraphy, fractures or coherent erosion.", "The inspected grazing five-tap sky estimate is about 3.30% below the dense reference (reference/five is 1.0341299). That small deficit cannot by itself explain very dark ground. Actual shadow components and display toe must be measured before increasing fill.", "Atmospheric multiple scattering, dense-cloud multiple scattering and sky/ground illumination of clouds are separate systems. One cannot stand in for another.", "The latest cirrus and natural cliff additions compile and pass structural checks but fail visual review. Both remain off and unselected.", "A 63.21 FPS average in one view is insufficient: its 18.22 ms p95 misses the 16.67 ms target, and the full route remains unaccepted." ], "contracts": [ { "name": "One world and Sun", "requirement": "All terrain, clouds, shadows and air use the same physical position, local vertical, solar direction, radius and time. Document any authored-to-curved coordinate transform.", "currentGap": "The dense far-Sun oracle follows the current straight authored-frame ray. It proves quadrature for that model, not curved planetary transport.", "test": "Compare flat and curved rays at near horizon, ground and high altitude with an independent path definition." }, { "name": "One energy convention", "requirement": "Radiance, irradiance, phase functions, BRDF factors and exposure use compatible units and normalization.", "currentGap": "Normalized presentation units and heuristic cloud/ground fills coexist with physical components.", "test": "Gray card, neutral sphere, homogeneous cloud slab and a common Sun/sky environment; inspect pre-tone component values." }, { "name": "Shared visibility", "requirement": "The same weather and terrain block sunlight consistently for surfaces, clouds and air, including off-screen casters.", "currentGap": "Several separately sampled visibility/cache paths and a proven far-Sun alias remain.", "test": "Move one cloud across Sun, terrain and camera; compare shadow, shaft and cloud-light responses from the same epoch." }, { "name": "Ordered media", "requirement": "Attenuation and in-scattering are applied along the actual ray segments, without double-counting clear air or using an inappropriate average depth.", "currentGap": "Cloud/air composition uses an opacity-weighted distance approximation that needs a bounded multi-layer comparison.", "test": "Two separated cloud slabs over terrain, then an in-cloud view, compared to joint ray integration." }, { "name": "Shared surface definition", "requirement": "Geometry, shading normals, material coordinates, POM, placement, navigation and shadows agree at every patch and LOD boundary.", "currentGap": "R32 repairs tested height contact accuracy; a visible material/shading seam and final motion checks remain.", "test": "Boundary scanlines and moving-camera probes through fine/local/far fields and triplanar/stochastic cells." }, { "name": "One temporal state", "requirement": "Density, lighting caches, motion vectors and exposure correspond to the same simulation time; history is rejected when it ceases to represent the current signal.", "currentGap": "Rigid advection is supported, but growth/deformation and weighted cloud depth remain approximate. Finite-shutter integration is unvalidated.", "test": "Frozen-camera evolving weather, moving-camera frozen weather and both together, against full-trace shutter-integrated references." }, { "name": "One color pipeline", "requirement": "Assets are decoded once, lighting stays linear, exposure/white balance are explicit, and display transfer is applied once.", "currentGap": "Transforms exist but the full camera/lighting pipeline is not calibrated to a photographic reference.", "test": "Gray/color/highlight ramps plus native render versus decoded film; retain unclipped HDR diagnostics." } ], "priorities": [ { "order": 1, "name": "Remove visible correctness defects", "status": "in progress", "work": [ "Retain the verified microgravel normal correction. Review residual foreground alignment near the slab and remaining shadow artifacts in native stills and motion.", "Resolve the authored/curved far-Sun contract; implement a stable shared optical-depth solution against a converged reference.", "Keep optional rejected features disabled." ], "exit": "Seam absent in paired native still and motion; low-Sun flashes eliminated without hiding them in history or grading." }, { "order": 2, "name": "Calibrate the complete light and camera chain", "status": "planned; component studies exist", "work": [ "Create a shared Sun/sky/gray-card/cloud-slab reference rig.", "Attribute grazing dark terrain to actual direct, sky, bounce, AO and tone components.", "Compare the current HG cloud material against a physically appropriate droplet model offline without violating the trained renderer's contract." ], "exit": "Documented units, altitude and visibility conventions; bounded errors and plausible photographed light ratios under fixed exposure." }, { "order": 3, "name": "Rebuild the dominant visible forms", "status": "research complete enough to scope; new design required", "work": [ "Author a connected sandstone landform with coherent layers, joints, ledges, undercuts and talus; validate clay silhouette first.", "Separate landform, metre-scale fractures and grain-scale material features.", "Improve cloud billow hierarchy, base organization, weather population and evolution using native 3D source/simulation rather than arbitrary surface noise." ], "exit": "Ground/aerial/grazing silhouettes and scale read correctly before materials or grading are used to enhance them." }, { "order": 4, "name": "Make surface and environmental response coherent", "status": "planned", "work": [ "Repair stochastic/triplanar/POM continuity and preserve material contrast.", "Correlate albedo, roughness, normal and displacement with geology; filter normal variance.", "Add or improve directional sky visibility and local ground interreflection only where matched references show the deficit." ], "exit": "No cells/seams/shimmer; stable daylight and low-Sun appearance with plausible contacts and shadow color." }, { "order": 5, "name": "Validate layered atmosphere and conditional weather", "status": "planned after core transport", "work": [ "Quantify air/cloud composition and altitude errors.", "Implement local dust/haze, shafts or ice-cloud layers only for a defined weather state and with consistent lighting.", "Keep rare halos, rain, lightning, heat shimmer and night optics scoped to shots that need them." ], "exit": "Weather layers add the intended physical cue without flattening the image or masking terrain defects." }, { "order": 6, "name": "Finish photographic motion and release", "status": "open", "work": [ "Use a shutter-integrated reference to select motion treatment.", "Validate full-resolution cloud reconstruction, terrain shading and camera controls in motion.", "Meet six-view and full-flight performance gates; render/decode the 4500-frame film and publish the accepted browser build." ], "exit": "Native 1080p60 on the RTX5080 with stable motion and no visible transport/reconstruction defects; verified local and public viewing links." } ], "budget": { "target": "1920x1080 at sustained 60 FPS on RTX 5080", "frameCeilingMs": 16.6667, "preferredMedianMs": 14.5, "proposedAllocationMs": { "terrainAndMaterials": 4, "cloudTransportAndReconstruction": 7, "atmosphereAndCoupledShadows": 2, "cameraAndComposite": 1, "reserveInsideMedian": 0.5 }, "warning": "These are planning allocations, not measurements or promises. Measure the complete frame, cache updates and all temporal phases. CPU preparation and submission must also fit the delivery cadence; they are not simply added to overlapping GPU stages.", "currentEvidence": "R32 ground only: 63.21 FPS average, 15.14 ms median, 18.22 ms p95. Aerial R32 cost and final flight acceptance are unmeasured. Uncontended measurements only; functional captures under known load carry no timing claim.", "strategy": "Spend memory and precomputation on stable data, share transport queries across pixels, and use selective reconstruction. A higher resolution setting is accepted only when it improves native appearance and fits complete-frame cost. Browser implementation choices must match actual WebGPU capabilities." }, "validation": [ "Record exact source/options/camera/weather/Sun/display for every comparison; change one relevant factor when attributing a defect.", "Use real photographs for appearance, generated art for direction and independent converged rendering for transport correctness; label each reference category.", "Inspect native 1920x1080 images and local crops, then actual-speed video. Numerical averages cannot hide localized seams, flashes or ghosting.", "Use below-cloud, cloud-edge, in-cloud, above-cloud, ground-grazing, backlit and low-Sun cases; include changing density, camera motion and source-cache updates.", "Require appropriate physical model, implementation correctness, visual quality, temporal quality and complete-frame cost separately. A pass in one is not a pass in the others." ], "conditionalCoverage": [ { "effect": "Ecology, sparse vegetation and biological soil crust", "trigger": "A vegetated Colorado Plateau habitat rather than a deliberately barren rock/sand basin.", "current": "No coherent ecology model is established by this audit; bare terrain is an explicit scene choice, not a universal property of a desert.", "solution": "Choose one habitat/reference first. If needed, distribute appropriate sparse plants and crust using soil, drainage and exposure masks, with scale-correct geometry, leaf transmission and contact shadows. Do not scatter generic shrubs uniformly.", "validation": "Compare the selected habitat's density, clustering and ground coverage to real photographs; test silhouettes and translucency under backlight.", "source": "https://www.nps.gov/cany/learn/nature/plants.htm" }, { "effect": "Wet ground, puddles and runoff", "trigger": "Rain or a recently wet surface is part of the selected weather sequence.", "current": "Dry material baseline; no coupled wetness state is accepted.", "solution": "Use a shared water/deposition state for absorption, surface-film reflection, roughness, puddle depth and drying. Wetness is not just lowering roughness everywhere.", "validation": "Dry-to-wet material samples under matched light, shallow-water reflection against the same sky/cloud scene and temporally coherent drainage.", "source": "https://diglib.eg.org/items/67f592b4-f58f-4893-8f63-2cbd81534558" }, { "effect": "Rain, virga, splashes and storm electrical light", "trigger": "A precipitating or thunderstorm state, rather than the present dry cumulus baseline.", "current": "Not implemented or validated for this scene.", "solution": "Model precipitation volume and fall/advection first; render streaks using shutter and lighting, preserve distant extinction and depth occlusion, and couple landing rain to wetness. Lightning needs illumination of both the cloud volume and landscape, not only an emissive line.", "validation": "Fixed rain-rate/camera exposure tests, cloud-base continuity, rain behind/over terrain, and any flash's full-scene energy/exposure response.", "source": "https://www1.cs.columbia.edu/CAVE/publications/pdfs/Garg_TOG06.pdf" }, { "effect": "Atmospheric refraction and heat shimmer", "trigger": "A close Sun/horizon shot or hot near-ground view where the chosen atmosphere produces a visible refractive effect.", "current": "Current clear-air transport primarily models extinction and scattering; a refractive gradient is not established.", "solution": "For a required shot, compare a stratified refractive-index ray reference, then approximate the bounded angular displacement with depth and distance dependence. Keep shimmer separate from camera shake and geometry/reconstruction artifacts.", "validation": "Horizon timing/solar shape and a fixed distant edge over a heated surface; no distortion of the near foreground or unrelated sky.", "source": "https://cloudatlas.wmo.int/en/general-classification-of-meteors-photometeors.html" }, { "effect": "Halos, coronas, glories, rainbows and iridescence", "trigger": "The particle material/size, Sun-view angle and weather state support the specific phenomenon.", "current": "Not part of the accepted HG cloud material; absence is not a defect in every shot.", "solution": "Use the appropriate angular/spectral droplet or ice scattering reference and preserve the transport model contract. Add only the phenomenon demonstrated by that reference, with correct angle and occlusion.", "validation": "Angular sweeps against a physical phase-function reference; distinguish atmospheric optics from lens flare.", "source": "https://cloudatlas.wmo.int/en/general-classification-of-meteors-photometeors.html" }, { "effect": "Moon, stars, airglow and night exposure", "trigger": "The film continues into an explicitly specified night rather than ending at twilight.", "current": "No complete calibrated night-lighting solution is claimed. A nonzero atmosphere floor is not a physical night sky.", "solution": "Add a coherent lunar/stellar sky and ground/cloud illumination with astronomical orientation, lunar phase and exposure appropriate to the intended camera. Preserve dark adaptation or manual exposure as an explicit camera choice.", "validation": "Sky and ground brightness ratios, star/Moon angular scale and cloud occlusion under one exposure; no invented daylight fill.", "source": "https://pbr-book.org/4ed/Cameras_and_Film/Film_and_Imaging" }, { "effect": "Polarization and full spectral transport", "trigger": "Matching a polarized-lens photograph or a chromatic atmospheric/material effect that the RGB reference cannot reproduce adequately.", "current": "RGB transport and unpolarized imaging are the working approximation.", "solution": "Use spectral/polarimetric offline comparisons to establish the error first. Add a targeted runtime approximation only when the mismatch is visible; a full spectral runtime is a separate architecture and cost decision.", "validation": "Same scene, camera and illuminant with/without the relevant filter/material; assess hue and angular behavior before grading.", "source": "https://pbr-book.org/4ed/Cameras_and_Film/Film_and_Imaging" } ], "referenceBoard": [ { "title": "Canyonlands geology and photographed strata", "kind": "real-world geology and photograph", "url": "https://home.nps.gov/cany/learn/nature/geologicformations.htm", "use": "Coherent sedimentary layers, differential erosion, ledges and whole-landscape organization. Photo credit NPS/Rhodes Smartt; exposure and lens metadata are not supplied here." }, { "title": "WMO: Cumulus mediocris from below, within and above", "kind": "observational cloud reference", "url": "https://cloudatlas.wmo.int/en/cumulus-mediocris.html", "use": "Cloud base, billow organization and aircraft-view morphology. This is morphological guidance, not a calibrated radiance reference." }, { "title": "Bruneton atmospheric-scattering implementation and validation", "kind": "physical reference implementation", "url": "https://ebruneton.github.io/precomputed_atmospheric_scattering/", "use": "Independent atmosphere/units/altitude comparisons." }, { "title": "PBRT image formation", "kind": "physical rendering reference", "url": "https://pbr-book.org/4ed/Cameras_and_Film", "use": "Camera, shutter, sensor and color calibration." } ], "updated": "2026-11-06T19:46:22.019Z" }