Uploaded September 2025 | Updated September 2026, 2 hours ago

---

Final simulation results for the South Tower (the tower that collapsed first chronologically). High-resolution physics simulation with original audio synthesized from the simulation output. Part 2 will cover the North Tower. This is an independent technical reconstruction and not part of any official investigation.

01:15 NBC Woolworth Building Statcam
02:13 Floor Failures
03:10 Distant View from West
04:10 Cut Through Both Towers
05:06 Marriott WTC-3 Close
05:52 Real Cam: 2008 Anonymous Release
06:38 Aerial View from Above (Wide Shot)
07:38 Following the Top
08:14 View from North Tower onto South Tower
08:38 View from South Tower onto North Tower
08:56 Aerial View of Plaza
09:16 Plaza at Eye Level
09:40 Underground / Mall

The simulation reconstructs the South Tower collapse after years of iterative development, verification and open-source tooling. Structural geometry was built from public imagery, plans and measurements where feasible and simplified when necessary to keep computation tractable. Core and perimeter columns, floor trusses, slabs and façade panels are represented as discrete components with explicit connections; steel is modelled with elastic–plastic behaviour, including limited strain-rate effects where practical. Connections are implemented with fixed constraints that evaluate each degree of freedom individually and can slip or rupture when thresholds are exceeded; spring-type constraints introduce an intermediate plastic phase that permits pre-failure yielding before final rupture.

The collapse trigger is a localized weakening of elements around the impact region, implemented as a spatially decaying gradient that reduces element strength and stiffness with distance from the centre of impact. In addition, a small number of core columns that engineering judgment indicated had suffered total failure after the impact were removed from the model; these removals are few and, taken alone, would not have produced global collapse. Trusses and façade elements in the affected area were also weakened. After this prepared initial state no further manual removals, scripted impulses or ad-hoc forces were applied: the sequence unfolds under gravity, inertia, contact and material response.

No soft-body dynamics were used; the solver is a rigid-body, constraint-based system with tuned contact resolution, time integration and collision handling to maintain stability at high event rates. Spring constraints emulate plastic deformation phases and the fracture model disconnects constraints or produces fragments when stress or energy criteria are exceeded, allowing progressive redistribution of loads and collapse propagation. Grouping and chunking strategies were used to limit memory pressure during long runs, and all key parameters and runs were logged.

All simulation states were cached as vertex/centroid animations for playback, rendering and post-processing. The audio track was synthesized directly from these cached states by converting element-centroid motion into sound using a collision-grain rendering technique: impact peaks detected in grouped signals trigger short grains that combine pitched carriers mapped from spatial grouping with band-limited noise bursts for high-frequency detail. Grains are windowed, optionally filtered, spatialised with distance-dependent attenuation and stereo panning, temporally jittered to desynchronize low frame-rate sampling, and resampled to audio rate. The resulting soundscape is emergent from the physical simulation and not Foley or stock recordings.

Multiple parametric runs explored sensitivity to the weakening gradient, material strength, connection properties, damping and the particular set of removed columns; selected outcomes were compared qualitatively with available imagery. The model contains simplifying assumptions: geometry, connection detail and constitutive laws are approximations, and thermal weakening is represented phenomenologically as a strength/stiffness gradient rather than via a fully coupled thermo-structural transient. The aim is to present a physically plausible collapse pathway under stated assumptions, not to claim a single definitive reconstruction.

This release covers Part 1 (South Tower); a separate simulation of the North Tower will follow. The project is privately conducted without external funding or institutional backing.

3D software: Blender — blender.org
Simulation: Bullet Constraints Builder — github.com/KaiKostack/bullet-constraints-builder
Fracture Modifier (custom builds supported) — blenderartists.org/forum/showthread.php?343637-Custom-Build-Blender-Fracture-Modifier

The sound pipeline adapts concepts from prior work in physically based contact and granular synthesis (O’Brien 2002; van den Doel et al. 2001; Roads/Xenakis/Truax; Raghuvanshi et al.) into a custom implementation that links structural collapse data directly to audio output.