Skip to main content
TWIXELS
TWIXELS NEURAL CAPABILITIES // SUITE v1.4

Engineered for Sub-Millimeter Precision & Instant Spatial Ingestion.

Explore the proprietary algorithmic capabilities powering Twixels’ zero-hardware WebAR, neural geometry reconstruction, and dynamic edge-cloud spatial computing architecture.

INGEST SPEED140ms / cycle
TRIANGLE DECIMATION94.2% retention
WEBAR ACCELERATIONWEBGPU // WASM
6-DOF TRACKING DRIFT<0.02mm jitter limit
MODULE MANIFEST

Foundational Architecture Modules

LATENCY BUDGET < 16.6ms PER FRAME
CAPABILITY 01

Neural Surface & PBR Reconstruction

NEURAL GAUSSIAN

Generates production-grade metric 3D primitives from unstructured multi-view captures or single-perspective inputs in under 4 seconds with automated microfacet geometry de-aliasing.

Interactive Pipeline ReconstructionStage: Metric Mesh Topology
Glowing wireframe metric mesh of an engine turbine with cyan vertices and telemetry markers
VERTICES: 1,842,910DENSITY: 0.12mm ISO
Normal Map Radiance Synthesis
Roughness / Specular Inversion
Metalness Channel Isolation
Subsurface Ambient Occlusion
CAPABILITY 02

Zero-Hardware WebAR Runtime

BROWSER-NATIVE

Instant sub-millimeter plane anchoring and full 6-DOF tracking inside standard mobile browsers (Safari, Chrome) with zero app download, zero spatial headset, and absolute zero friction.

TARGET: HORIZONTAL PLANE
ANCHOR LOCKED // 60 FPS
Payload < 450 KBSLAM & Gyro FusionQuickLook / WebXR Dual EngineDepth Occlusion Culling
CAPABILITY 03

Distributed Spatial Compute

HYBRID RUNTIME

Workload arbitration between device-side neural hardware (Apple Neural Engine / WebGPU) and cloud ray-tracing clusters guarantees sub-16ms telemetry even on low-power mobile nodes.

Dynamic Load BalancingAUTOMATED ARBITRATION
Client Node: 18%
Cloud Mesh: 82%
SYNCHRONIZATION JITTER< 1.4ms (QUIC UDP)
CAPABILITY 04

Industrial CAD & LiDAR Digital Twins

B-REP & MESH

Direct volumetric ingestion of parametric STEP, IGES, and massive LiDAR aerial point clouds. Intelligent quad-dominant polygon reduction condenses multi-million facet models for instantaneous web delivery.

ORIGINAL CAD12,400,000Raw Polygons
-98.7% REDUCTION
TWIXELS WEBGPU150,0000% Visible Artifacts
Native Parsers:STEP / IGESGLTF 2.0 / USDZLAS / LAZ LiDARIFC BIM
EMPIRICAL BENCHMARKS

Architectural Supremacy at Scale

Measured against industry standard photogrammetry clusters and legacy spatial application paradigms.

Evaluation Metric
TWIXELS NEURAL CORE
Legacy PhotogrammetryBulky Native App Solutions
Model Generation Time
Input upload to interactive runtime
3.8 seconds45 to 120 minsN/A (Pre-rendered)
Client Payload Size
Network overhead transferred to client
< 450 KB (WASM core)> 85 MB per scene1.2 GB+ native binary
Surface Anchoring Drift
Rotational and translational slippage
< 0.02 mm / min> 3.5 mm / min0.10 mm / min
Device Compatibility
Target execution environment requirements
100% Modern Web BrowsersHigh-End CUDA WorkstationsDedicated Headsets / App Stores
DEVELOPER TOOLCHAIN

Embed Spatial Ingestion with Four Lines of Code.

The @twixels/spatial-core package mounts directly inside modern React, Vue, or headless TypeScript stacks. Configure hardware thresholds dynamically.

spatial-client.tsx
import { TwixelsEngine, SpatialViewport } from '@twixels/spatial-core';

export const IndustrialSpatialViewer = () => {
  const engine = new TwixelsEngine({
    apiKey: process.env.TWIXELS_SPATIAL_KEY,
    webgpu: true,
    autoDecimation: {
      enabled: true,
      budgetTriangles: 150000,
    },
    tracking: {
      mode: 'zero-drift-slam',
      subMillimeterThreshold: 0.02
    }
  });

  return <SpatialViewport runtime={engine} assetId="cad-9982a" />;
};
DEPLOYMENT READINESS

Ready to Integrate Twixels Capabilities into Your Platform?

Connect our WebGPU zero-app runtime with your current CAD repositories, e-commerce catalog, or spatial field-engineering tools.