OpenLiDARViewer

A browser-native LiDAR and point-cloud viewer for fast local inspection, 3D navigation, measurement, and terrain analysis. Local-first, cited, honest about what it can't tell you.

Try it in 10 seconds

No install, no account, no upload. Open app.openlidarviewer.org, then drag a .las, .laz, or .copc.laz file (or paste a remote COPC / ept.json URL) onto the page. You're navigating the cloud in your browser, and the file never leaves your device.

New here? The User Guide covers opening and measuring a scan. It also covers terrain analysis, comparing two scans and sharing your work. Full documentation, the format matrix, and the scientific validation record live at the docs site.

Overview

OpenLiDARViewer opens LiDAR and point-cloud datasets straight in the browser. You inspect a scan, navigate it in 3D, change how it is coloured, measure, run terrain analysis and export the results, with no desktop GIS to set up. Files are read and rendered locally, so there is no server to upload to.

It is a viewer and an inspection tool, not a GIS or a survey-grade processing suite. Every result states its coverage, its method and its uncertainty, and a terrain or contour export is gated on evidence rather than produced on request.

Features

• Inspect point-cloud datasets in a modern web interface, with nothing to install and nothing uploaded.
• Open drone LiDAR (LAS, LAZ) and terrestrial laser-scanner data (E57, PTX, PTS). iPhone and mobile scan exports (PLY, OBJ, GLB/GLTF, XYZ, CSV) and Point Cloud Library (PCD) files open too.
• Stream large COPC and EPT datasets progressively, octree node by octree node, with bounded memory and no full-file load.
• Navigate game-style: Orbit, Walk, Fly, and Pan modes with WASD and mouse-look, plus Top / Iso / Oblique / Planar camera presets.
• Measure distance, polyline, area, height, angle, slope, cross-section profile, and volume cut/fill, with editable points and JSON session export/import.
• Run a confidence-aware DTM and contour pipeline: ground classification, gridded DTM with hold-out RMSE, surface models, and evidence-gated contour and DEM export.
• Click any point to read its coordinates, intensity, classification, GPS time and colour. The Scan Intelligence panel summarises the dataset, and findings can be annotated and composed into multi-page PDF reports.
• Switch themes (Dark / Light / High-contrast), drive everything from a command palette (Cmd-K), and work one-handed on mobile.

OpenLiDARViewer does not claim survey-grade measurement or support for every LiDAR format. Capabilities are described honestly (see Limitations and the details below).

Core viewer

• Browser-based point-cloud visualization, no install
• Local-first scan inspection, nothing is uploaded
• WebGPU rendering with an automatic WebGL 2 fallback
• Import: LAS, LAZ, E57, PLY, OBJ, GLB, GLTF, XYZ, CSV, PCD, PTX, PTS
• Export: LAS (1.2 / 1.4), PLY, OBJ, XYZ, CSV, and PNG snapshots
• Budget-aware fast loading of large LAS/LAZ surveys: header preflight, stride decoding, a memory-safety guard, staged progress, and a load that can be cancelled mid-flight
• Chunked, bounded-memory reading of large text point clouds (XYZ, CSV, PTS); a large survey reduces in density instead of crashing on weak devices
• A universal file-open summary and clear, categorised load-error messages
• A coordinate bridge that keeps large georeferenced (UTM-scale) coordinates numerically stable

Streaming

• A COPC .copc.laz file, on disk or hosted at a URL, opens through progressive, octree-based, view-dependent streaming with worker-based decoding and bounded memory, never a full-file load. A remote scan opens from the start screen's open-from-URL field or a shareable ?copc=<url> deep link
• EPT (Entwine Point Tile) streaming: local and remote, binary and laszip tiles
• 3D Tiles / .pnts: a single .pnts tile opens as a point cloud. It is detected by its magic bytes, decoded from uncompressed or quantised positions with colour, and placed by its RTC_CENTER
• 3D Tiles / tileset.json: a 3D Tiles 1.0 or 1.1 tileset whose content is PNTS opens from a URL and streams the way COPC and EPT do. The scheduler culls against the camera, selects what fits the point budget, and fetches and decodes tile bodies as they are needed; each tile is placed by its own cumulative transform, after its RTC_CENTER and before the render origin, in float64. A streamed tileset reports no source point total, because a tileset.json never states one and the per-tile figures are decode-admission estimates rather than counts. A tileset whose hierarchy is implicit, a quadtree or octree of subtree files rather than a written-out tree, opens too: it is expanded to the equivalent explicit tileset before parsing, so the same refusals apply. Both refinement modes stream: ADD draws every resident node, and REPLACE swaps a parent for its children atomically once all of them are resident, with no doubled geometry and no hole. A tile whose content is a nested external tileset.json is followed and spliced in as a child, so a set split across several documents opens as one scene. Both the single content and the 3D Tiles 1.1 contents[] array are read, and a tile whose entries are all point clouds streams every one. Mesh content (B3DM, I3DM, CMPT, glTF) and Draco are refused by name, because point streaming is the whole of the subset. See docs/supported-formats.md for the limits of the subset
• A curated catalog of 12 hand-vetted public COPC / EPT datasets

See docs/streaming.md and docs/copc.md.

Navigation & camera

Game-like navigation lets a scan be explored like a 3D environment.

• Orbit, Walk, Fly, and Pan (hand-tool) modes with WASD movement and mouse-look
• Smart camera presets (Top, Iso, Oblique, Planar): one-click jumps that frame the cloud from a known angle
• A triangular nav widget that surfaces the current mode and a centre Reset
• Saved, renamable camera views for repeatable inspection
• Shareable view links: Copy view link reproduces the current view (camera, colour mode, point sizing); no scan data is shared, the recipient still needs the same file
• Movement speed scales with the size of the loaded scan, so the controls feel right for a small room or a kilometre-wide survey

Control	Action
W / A / S / D	Move through the scan
Mouse	Look around (click the scan to capture the cursor)
Shift	Move faster
Space	Move up
C / Ctrl	Move down
Esc	Release the cursor
R	Reset / re-frame the view
F	Focus on the point under the cursor
1 / 2 / 3 / 4	Orbit / Walk / Fly / Pan mode
G	Toggle the Pan (hand) tool
Middle-drag	Pan the view in any mode
Double-click	Fly to the clicked point

Orbit suits inspecting an area from the outside, Pan is the 1:1 hand tool, Walk suits interiors and street-level scans, and Fly suits drone LiDAR, terrain, and wide-area scans. Full detail is in docs/navigation.md.

Rendering

Rendering is tuned so a point cloud reads as a 3D surface, not a flat wash of dots.

• Eye Dome Lighting (EDL): screen-space depth shading that darkens depth discontinuities so edges, ridges, and near/far structure become legible. It runs as a post-processing pass targeting both the WebGPU and WebGL 2 backends from one node graph, with strength, radius, and three named presets (Subtle / Balanced / Inspection). On by default on desktop WebGPU, off by default on the WebGL 2 fallback and on mobile, where it can still be switched on
• Hillshade relief overlay from the terrain surface model for topographic readability
• Soft splat rendering: Classic round points, Soft splats, or Inspection mode with density-aware radius
• Adaptive or fixed point sizing, round antialiased points, and a Detail control that shows an honest shown / total count
• Height, intensity, classification, RGB, and surface-normal colour modes, picked automatically per file
• Percentile-clipped height mode with a 5/95 default and a Turbo perceptual palette
• HDR sky presets: Studio Dark, Blueprint, Survey Light, Terrain, Black

Visuals Studio

• A unified Inspector section with three chip rails (Colour Mode, RGB Preset, Sky / EDL) that re-style the scan without leaving the panel
• RGB appearance presets (Photoreal, Drone RGB, Mobile LiDAR, Infrastructure), each tuning gamma, contrast, saturation, and exposure
• White balance (temperature + tint) with an Auto-balance assist, gated to streaming COPC where it can sample residency
• Patch view at the inspector cursor: a KNN-based tangent-plane projection of the point's neighbourhood with a photometric witness panel

Measurement & analysis

Open the Measure tool, pick a kind from the toolbar, and place points directly on the scan. Seven tools are available.

Tool	What it measures
Distance	Straight-line distance between a pair of points
Polyline	Total length of a multi-segment path
Area	Polygon area, both true in-plane and horizontal (map-projected)
Height	Vertical difference between a pair of points
Angle	The angle at a vertex between two arms
Slope	Rise, run, slope angle, and grade percentage between a pair of points
Profile	Cross-section between a pair of points: 3D length, horizontal distance, vertical drop, and grade

• Every measurement is editable: drag a point, undo the last point while placing, rename, or clear. Placed measurements list in a compact panel and persist for the session. One toggle switches all readouts between metric and imperial. The set exports to a JSON session file and re-imports later
• Cross-section profile renders a height-vs-distance chart strip under the row, resizable from a default 140 px out to 360 px so the curve reads at deliverable size
• Volume (cut / fill) against a polygon or 3D lasso, with NaN / degenerate / self-intersection guards and a streaming-resident caveat when nodes are still loading
• Classification editor: paint a class id over a lassoed selection and write the result back to LAS
• Density heatmap overlay for coverage QA, box clipping / slicing for interactive cross-cuts, and measurement chains that combine placed measurements as sum / difference / ratio
• A Scan Intelligence panel with point count, dimensions, density, spacing, attributes, and an Advanced report of integrity diagnostics
• A Dataset Intelligence card (header-derived Point Density, Terrain Complexity, Ground Visibility, Streaming Coverage, Terrain Confidence) that leaves a row blank rather than fabricating a bucket when no signal is available
• Point inspection: click a point to read its coordinates and attributes (LAS return number, point source ID, GPS time, and UTM + lat/lon when a CRS is known), with one-click copy, or hover with the live probe for a click-free readout
• Capture provenance from LAS/LAZ and E57 headers (sensor, source software, date), shown in the Scan Report when the file carries it

Measurement is meant for visual inspection and research, not survey-grade use. OLV makes no survey-grade claim; if accuracy matters, check against ground control using your own procedures.

Terrain Intelligence & Contour Studio

OpenLiDARViewer ships a terrain analysis stack under src/terrain/. Shared type contracts (TerrainContracts.ts) give every stage one honesty envelope: a coverage mode (full / resident-only / sampled), the source and analyzed point counts, a 0-100 confidence value, and ordered warnings, so an analyser never implies full-cloud certainty when only resident streaming nodes were walked.

The lightest surface is the Dataset Intelligence card in the Inspector: header-derived, informational, and it does not perform ground classification. The main capability is the confidence-aware DTM and contour pipeline (src/terrain/contour/, ground/, surface/) surfaced through the Analyse panel: ground classification, a gridded DTM with per-cell confidence and hold-out RMSE validation, a 0-100 heuristic terrain readiness index (workflow-readiness weights, not an external accuracy calibration), surface models (DSM, canopy height, slope, multi-directional hillshade), a single top-level Terrain Assessment verdict, evidence-graded contour export (GeoJSON / SVG / DXF), a printable map sheet, and a georeferenced DEM package (ASCII Grid + GeoTIFF). Ground classification is a heuristic: its output is derived, not survey-grade. A DTM quality gate governs whether terrain-product export is enabled, and per-cell confidence is calibrated against measured hold-out error, not asserted. Treat terrain products and DEM exports as export-ready only when the Terrain Assessment reads Good, and as preview otherwise.

Contour Studio is the post-analysis step that turns an analysed scan into a contour deliverable, kept out of the Analyse panel so map-making doesn't crowd the terrain work. You pick a purpose: one of the four named ones (Engineering Plan, Survey Review, Terrain Research, Presentation Map) or Custom. A purpose only bundles presentation defaults and can never raise a claim. Analytical contours are the exact isolines of the grid, while cartographic contours are generalised for legibility, reference the analytical geometry's hash, and are never labelled exact. Every export routes through one evidence gate that can only downgrade (validated, exploratory, or blocked): a blocked product returns a diagnostic instead of a polished file, an exploratory one is watermarked, and the permit decision is stamped into each artifact's provenance. Exports cover contour vectors (GeoJSON, DXF, SVG), a map-sheet PDF, a DEM raster package, a terrain intelligence report, and a complete ZIP with a SHA256SUMS manifest. Validation is internal hold-out only: nothing is survey-grade, and no output asserts certification.

See docs/terrain-intelligence.md, docs/validation/terrain-validation-matrix.md, and docs/contour-studio.md.

Annotation, sessions & reporting

• Annotations are categorised, titled markers with notes. You can browse and search them, and each one captures the camera viewpoint; undo and redo cover every edit. The panel and the PDF report open with a grouping summary (totals, per-category counts, and how many areas the notes fall across)
• Inspection sessions: export measurements, annotations, and named views to one JSON file and reload them later
• The workflow recorder records and replays .olvworkflow files of camera moves and tool actions. Its settings cover the file format, save destination and start/stop shortcut. Replay has its own speed and loop options, recording can start after a countdown, and you choose which action families are captured. It records actions only, never scan data, so a recipient needs the same scan open to replay
• Multi-page PDF technical reports: three built-in templates (Survey Summary, Technical Report, Scan QA) with branding and unit-system awareness
• Visual Export Studio: orthographic RGB, height map, intensity, classification, depth, normal, and contour map exports
• Screenshot export that burns in placed measurements and annotations as inspection evidence

Interface & accessibility

• Theme system (Dark, Light, High-contrast) with a persisted preference
• A colourblind-safe (Okabe-Ito) classification palette, toggled from the Classes panel, so ground, vegetation, buildings, and water stay distinguishable under the common colour-vision deficiencies
• Command palette (Cmd-K / Ctrl-K) for keyboard-first access to every tool, mode, theme, and export
• Searchable shortcut sheet (?) listing every keybinding, plus a built-in help overlay
• An onboarding tour through the empty state, tool dock, and Inspector
• A mobile touch model with twist + pinch + pan decomposition, sub-threshold dead zones to stop accidental wobble, and an opt-in 3-finger zoom
• A mobile Scan Intelligence bottom-sheet with peek + tap-to-toggle, and an overflow "More" disclosure on the tool dock so the primary row stays one-handed

Multi-scan & embed

• Open multiple scans as layers, or close the current scan from the tool dock to start fresh with another
• An embed mode for <iframe> use (?embed=1), with a validated postMessage bridge for host-page control
• Developer diagnostics: a live performance overlay (?debug=1) and a structured benchmark mode (?benchmark=1)

Screenshots

A 9.6M-point drone survey, height-colored, with the Scan Intelligence panel and Orbit / Walk / Fly navigation.	The measurement toolkit: here a distance between two picked points.
Inspecting a point: a glowing marker and a card with its real-world coordinates and attributes.	The Scan Intelligence panel lists point count, dimensions, density and spacing. Attributes and the Advanced report sit below.

More in docs/screenshots.md.

Formats & requirements

Import: LAS, LAZ, E57, PLY, OBJ, GLB, GLTF, XYZ, CSV, PCD, PTX, PTS. Export: LAS (1.2 / 1.4), PLY, OBJ, XYZ, CSV, and PNG snapshots.

For large datasets, stream COPC (.copc.laz) or EPT (ept.json), which load progressively with bounded memory. For lightweight sharing, use PLY or GLB.

Format matrix and compatibility notes

That covers iPhone and mobile scan exports (PLY, OBJ, GLB/GLTF; USDZ needs conversion first), terrestrial laser-scanner data in E57 (the ASTM E2807 format; a subset, read-tested against Trimble exports, with no conformance to that standard claimed) plus PTX and PTS, georeferenced drone LiDAR in LAS/LAZ, and PCD in all three encodings (ASCII, binary, binary-compressed). Large COPC and EPT datasets stream progressively, locally or over HTTP range requests from a URL, with bounded memory and no full-file load.

Format support depends on the device (browser memory, GPU capacity) and on the data: its size, how it was preprocessed, and how complete OLV's reader for that format is. The per-format detail, including scanner and app compatibility, is the format matrix in docs/supported-formats.md.

System requirements

OpenLiDARViewer runs in the browser and depends on modern GPU-accelerated web rendering. Performance varies with the dataset and the device. Use a modern Chromium-based browser (Chrome or Edge) with WebGL 2.0 support and hardware acceleration enabled. WebGPU is used automatically where it is available, including in Firefox and Safari on the platforms where they expose it, with the WebGL 2 fallback used elsewhere (see docs/developer-manual.md section 14 for the per-browser breakdown): a boot-and-render smoke (.github/workflows/browser-smoke.yml) and the full deterministic end-to-end suite run on both engines as an advisory cross-browser matrix, which no merge rule requires.

Component	Minimum	Recommended
CPU	Modern dual-core	Quad-core or better
RAM	8 GB	16 GB or more
GPU	Integrated GPU with WebGL 2.0	Dedicated GPU, or modern Apple Silicon / integrated GPU
Browser	WebGL 2.0 compatible	WebGL 2.0 and WebGPU-capable

Very large datasets are best handled as COPC or EPT; other very large formats may need downsampling or preprocessing. See docs/copc.md and docs/performance.md.

Mobile browser support

OpenLiDARViewer includes a mobile-friendly interface for opening compatible point-cloud and 3D scan files from phones and tablets. On mobile: files open from the device file picker or a cloud file provider, Scan Intelligence shows as a compact panel, navigation uses touch gestures, measurement uses tap-based point selection, and rendering defaults to a mobile-safe performance mode.

A typical workflow is to export a compatible scan from a mobile scanning app, save it to device storage or a cloud provider (such as iCloud Drive), open OpenLiDARViewer in a mobile browser, tap "Open scan from device," then inspect, measure, and export. For a practical test, capture a scan with an iPhone LiDAR app such as Polycam, Scaniverse or 3D Scanner App. Export it as GLTF/GLB, OBJ or PLY and open the file. Export formats, free-tier options, and pricing differ between apps and can change, so check each app's current help. Some formats may require a paid plan.

On mobile, the browser, GPU and memory set the ceiling, and file size and point count decide how close a scan gets to it. Very large datasets may require desktop hardware, downsampling, tiling, or optimized formats. All third-party product names are used only for descriptive compatibility. OpenLiDARViewer has no affiliation with Apple, Polycam or any other scanning-app maker, and none of them endorses or sponsors it. Full detail is in docs/mobile-browser-support.md.

Getting started

git clone https://github.com/Aurtechmx/openlidarviewer.git
cd openlidarviewer
npm install
npm run dev

Open the local URL it prints, then drop a scan onto the page or click a built-in sample. To build for static hosting (GitHub Pages, Netlify, or any CDN, since it is just files):

npm run build
npm run preview

Using the viewer

1. Open the app in a modern WebGL/WebGPU-capable browser.
2. Drop a compatible point-cloud file onto the page.
3. Choose a colour mode: Height, Intensity or Classification, or RGB and Normal where the file carries them.
4. Adjust point size and rendering detail.
5. Navigate with Orbit, Walk, or Fly mode.
6. Read the Scan Intelligence panel for dataset metadata and quality.
7. Measure distance, polyline, area, height, angle, slope, or cross-section profile inside the point cloud.
8. Annotate points of interest with categorised notes, and inspect or probe individual points.
9. Save viewpoints for repeated inspection.
10. Export a PNG snapshot, re-export the cloud as PLY, OBJ, XYZ, or CSV, or save the full working state as a .olvsession package.
11. Close the scan from the tool dock to return to the start and open another.

A fuller walkthrough is in docs/usage.md.

Recommended workflows

Each assumes a single drag-and-drop or URL open, with everything happening locally in the browser.

• Large streaming dataset review. Open COPC (.copc.laz) or EPT (ept.json), local file or remote URL. Navigate at interactive frame rates against datasets far larger than browser memory; the scheduler streams only what the current view needs.
• Inspection reporting. Annotate findings, measure distances / areas / slopes / angles / profiles, then export a multi-page PDF report (cover, dataset summary, embedded image exports, annotations, measurements, technical notes). Three templates and brand-aware accent + logo support.
• Terrain analysis. Export height maps from drone LiDAR with legend customisation and unit-system control, useful for slope review, elevation comparison, and quick topographic figures. Cross-section profiles report 3D length, horizontal distance, vertical drop, and grade across any two picked points.
• Classification QA. Export classification maps, toggle the colour mode to highlight specific classes, place annotations on misclassified regions, and round-trip the working state through .olvsession.
• Mobile scan review. Open lightweight datasets (.glb, .ply, .obj from Polycam, Scaniverse, or similar) on tablets or phones. The viewer adapts rendering detail and EDL defaults for weaker GPUs so a phone scan is readable from the first frame.

Under the hood

How it works

1. You load a dataset by dropping a file, or by clicking a built-in sample.
2. The format is detected from the file's magic bytes first, then its extension.
3. The file is parsed off the main thread, inside a Web Worker.
4. Point positions and attributes are decoded. Large georeferenced coordinates are recentered in double precision before the float32 downcast.
5. Clouds above the point budget are voxel-downsampled, and the Detail control shows the honest shown / total count.
6. The cloud renders through a WebGPU or WebGL 2 pipeline built on three.js; Eye Dome Lighting adds screen-space depth shading as a post-processing pass.
7. Color modes map height, intensity, classification, RGB, or surface-normal direction onto the points, sized adaptively with distance.
8. You explore with Orbit, Walk, or Fly navigation.
9. Scan Intelligence summarizes the dataset, and the measurement toolkit takes its measurements.
10. You save viewpoints and export snapshots, re-exported point data, or a JSON measurement session.

Technology stack

• TypeScript in strict mode, across the IO, model, and render layers
• three.js (three/webgpu), a WebGPU renderer with a WebGL 2 fallback
• A three/tsl node-graph post-processing pipeline (Eye Dome Lighting) that targets both backends from one shader description
• loaders.gl and laz-perf (WASM) for mesh and LAZ parsing, plus a from-scratch TypeScript E57 parser
• Vite for the build and dev server, with Web Worker and WASM handling
• Vitest and Playwright for unit and end-to-end tests
• A client-side, local-first pipeline with no backend

Architecture

OpenLiDARViewer is modular, with one file per format and one file per concern. Loading and parsing, the coordinate bridge and render-buffer generation each sit in their own modules. Color modes, navigation, measurement, Scan Intelligence and export are separate subsystems as well. See docs/architecture.md and the Developer Manual.

Performance notes

Point count, browser memory and GPU capability matter most for performance. Point size, rendering detail and color mode change the per-frame cost, and the file format and how the data was prepared change the load cost. A LAS/LAZ file is planned from its header before it is fully read: a cloud above the device's point budget, roughly 3M points on a capable desktop and less on weaker hardware, loads at reduced density (voxel-downsampled, or stride-decoded when far over budget so it is never fully decoded into memory; the source file is still read in once, while COPC, EPT, and the out-of-core path for a very large LAS or LAZ stream only the resident set), with a memory-safety guard, staged progress, and a cancellable load. The Detail readout always shows the honest shown / total count.

COPC streaming (local and remote) ships in v0.3.0 and is hardened across v0.3.1 / v0.3.3 with a view-dependent scheduler, hierarchy-aware eviction, a dispatch-pressure gate that bounds residency under 1B-synthetic-point stress, and trustworthy picking against actively-refining clouds. EPT joins COPC as a first-class peer in v0.3.3.

For real-world figures (a 9.6M-point drone LAZ survey and a 55K-point iPhone scan, both from one drag-and-drop) see docs/benchmarks.md, docs/performance.md, and docs/streaming.md.

Limitations

OpenLiDARViewer is a viewer, not a replacement for a GIS or a survey-grade processing tool. It leaves CRS reprojection to dedicated tools and analyses each scan in its own frame.

• Large files are limited by browser memory and GPU performance; some formats need preprocessing or conversion before they load, and format support is still evolving.
• Measurement is for visual inspection, not survey-grade use.
• Metric figures fail closed when the CRS linear unit is unconfirmed: the viewer reports extents in the source unit rather than fabricating metres.
• Coordinate reference system handling is basic; there is no cross-CRS reprojection.
• Classification visualization depends on attributes present in the file.
• Very large datasets stream as COPC (local or remote); other huge formats may still need tiling or downsampling.
• WebGPU feature support varies by browser, with the WebGL 2 fallback used otherwise. Eye Dome Lighting is a screen-space depth cue, not physically-based lighting, and is off by default on the WebGL 2 fallback and on mobile.

Full detail is in docs/limitations.md.

FAQ

Can I view LAS / LAZ / COPC files in the browser? Yes. Drag a .las, .laz, or .copc.laz onto app.openlidarviewer.org, or paste a remote COPC / ept.json URL. No install, no plugin.

Is my data uploaded anywhere? No. Files are read and rendered locally. The only network calls are for remote datasets you choose to open; your local files never leave your device.

What's the largest scan it can open? Most local files are bounded by browser memory and the GPU. A very large uncompressed LAS or chunked LAZ is the exception: it is indexed out of core into browser storage and streamed through the same scheduler, and reopening the same file reuses that index. A content hash authorises the reuse, so an edited file never gets a stale one. The index is a cache, not durable storage: the browser or you may clear it, and old ones are evicted under a size cap. Where storage is missing or too small, the file is refused with guidance rather than loaded whole.

For anything else too large, stream it as COPC or EPT, or convert it with PDAL or Entwine. Streaming loads only the set the camera needs.

Which formats are supported? For static loads, LAS / LAZ, PLY, XYZ / CSV, E57 and glTF / GLB; for streaming: COPC, EPT, a 3D Tiles PNTS tileset, and a very large uncompressed LAS or chunked LAZ indexed out of core. See Formats & requirements.

Is it survey-grade? No. Measurements and quality grades describe the data you loaded; they are not a survey-grade certification. Validate against ground control where accuracy matters.

Does it need WebGPU? No. WebGPU is the primary path and it falls back to WebGL 2 automatically.

Project & research

OpenLiDARViewer started as an experiment: how far can modern browser technology go in making LiDAR and point-cloud data easy to reach? It explores browser-native rendering, lightweight WebGL/WebGPU pipelines, human-centered interaction with 3D data, game-inspired navigation for technical inspection, and local-first workflows. It does not try to replace full GIS or survey-grade processing. The goal is a quick way to open a point cloud, look around it and measure it, and then present what you found. See docs/research-notes.md.

The current release is v0.7.0-alpha.1. The dated history is in CHANGELOG.md, and per-release records live in docs/releases/.

Help test OpenLiDARViewer

OpenLiDARViewer improves through feedback from people who work with point clouds day to day: GIS, drone mapping, terrain analysis, hydrology, surveying, web mapping. Open a workflow you already know, on the live demo or a local build. Use one of your own authorised files, compare the values you care about against the tool you normally trust (ArcGIS, CloudCompare and PDAL are the common ones), and say what worked, what failed, and what was unclear.

The quick report takes five to ten minutes. A longer comparison against a reference tool (metadata, CRS, units, elevations, measurements, terrain products) is optional. Participation is voluntary and unpaid. Please do not submit confidential, restricted, or personal information, and do not send source datasets you are not free to redistribute. A failed file, an unexpected warning, or one confusing screenshot is worth sending; negative results are the useful kind here. Email what you found to info@aurtech.mx.

Validation & reproducibility

For reviewers, and anyone who wants to check the claims above rather than take them on trust:

• REVIEWER_QUICKSTART.md: install and run the offline test suite from a clean clone in about two minutes.
• VALIDATION_REPORT_v0.7.0-alpha.1.md: what this release validates and what it does not. No product changed evidence level this cycle; the register holds 37 claims, 17 of them at E4 and none at E5. One measurement method changed: the interactive stockpile volume is integrated over an area grid, and VOL-STOCKPILE stays at E3. Two new claims were added for Terrain Flow Pulse's depression inventory work (TERRAIN-FLOW-PULSE at E3, TERRAIN-FLOW-DEPRESSION-INVENTORY at E2), and one new claim for Terrain Access's traversability screening (TERRAIN-ACCESS at E3, internal-oracle ceiling); no existing claim's evidence level changed.
• KNOWN_LIMITATIONS_v0.7.0-alpha.1.md: the documented limits of this release (no evidence promotion, a stockpile record that is not integrated beyond the toast, a registration stack that ships without a user path, touch gestures run in CI as synthesized events with no real-device check, one measurement figure with a known basis problem, no cross-CRS reprojection).
• REPRODUCIBILITY.md: the pinned toolchain and the steps to reproduce the build, tests, and reported figures.
• ARTIFACT_EVALUATION.md: how to evaluate the artifact without special hardware or private data.
• DATA_AVAILABILITY.md: where the test fixtures and streamed sample datasets come from, and how they are licensed.

Contributing

Contributions are welcome. See CONTRIBUTING.md, the security policy, and the code of conduct. The codebase is test-first (Vitest and Playwright), written in strict TypeScript, and organized around explicit module boundaries.

Acknowledgements

OpenLiDARViewer stands on a lot of open work, and we're grateful for it.

Built on three.js (rendering), loaders.gl (format parsing), proj4js (CRS transforms), pdf-lib (reports), and laz-perf (LAZ decoding). Full licenses in THIRD_PARTY_NOTICES.md.

The streamed sample datasets are limited to sources with a confirmed open licence: USGS 3DEP (public domain) and the swisstopo and GURS national programmes (via FLAI). Providers and terms are listed in docs/credits.md.

Format specifications OLV builds against (no conformance claimed): ASPRS (LAS/LAZ), the Khronos Group (glTF/GLB), ASTM (E57), and OGC / IOGP-EPSG (coordinate systems). Particular thanks to Howard Butler and Hobu, Inc., whose work on laz-perf, COPC, and Entwine this viewer relies on.

Citation & research collaboration

OpenLiDARViewer v0.6.7 and later is licensed under AGPL-3.0-only so its code and methods stay open, including when a modified version is offered over a network. The licence lets you use, study, change, and redistribute the software under those terms, and it says nothing about citation. Citation is a separate request: if the software, its validation framework, algorithms, or documentation helped your research, software, thesis, or commercial work, please cite it. GitHub's "Cite this repository" button reads CITATION.cff and gives you the current reference, including BibTeX.

Being cited is the minimum; working together is the better outcome. Independent reproduction, new surveyed-checkpoint datasets, terrain and contour cross-checks against a reference tool, and method comparisons feed the evidence record directly, and a result that holds or fails on your data tells us more than one that only holds on ours. If you have a dataset or a study in mind, open an issue or a discussion. The specifics, and what we can offer in return, are in docs/collaboration/RESEARCH_COLLABORATION.md.

License

OpenLiDARViewer v0.6.7 and later is licensed under the GNU Affero General Public License v3.0 only (AGPL-3.0-only). See LICENSE. Releases through v0.6.6 were distributed under the MIT License and remain available under those original terms.

OpenLiDARViewer is provided as is, without warranty of any kind; see sections 15 and 16 of the GNU AGPL.

Files you open are processed on your device and are not uploaded, and the app has no analytics or telemetry. See PRIVACY.md.

Commercial licensing may be available separately from Aur Technologies for eligible components and use cases such as closed-source embedding, OEM redistribution, or proprietary integration. See LICENSING.md, COMMERCIAL-LICENSING.md, and THIRD_PARTY_NOTICES.md.

If you use OpenLiDARViewer in research, a CITATION.cff is included. Developed by Aur Technologies (aurtech.mx).