---
title: "Scalable Spectral Packing: Pack Studio Technology"
url: "https://www.pack-studio.com/technology"
description: "Pack Studio runs on Scalable Spectral Packing (SSP), an Inkbit and MIT algorithm published at SIGGRAPH 2023. How it works, benchmarks and provenance."
updated: "2026-11-17"
type: "WebPage"
---

Technology /// Provenance

# Scalable Spectral Packing: The Algorithm Behind Pack Studio

**Peer-reviewed research, built in-house.**

Pack Studio's packing engine is Scalable Spectral Packing (SSP), developed by Inkbit and MIT researchers and published in ACM Transactions on Graphics, SIGGRAPH 2023[^1]. This page documents what it is, how it works and where it came from.

[Read the paper](https://doi.org/10.1145/3592126) · [Book a Demo](https://www.pack-studio.com/contact)

## Packing real parts is hard.

Placing many irregular 3D objects densely into a container is a classic hard problem in computational geometry. Exact solutions do not scale, and even good approximations are difficult when parts have arbitrary shapes and the pack must come apart cleanly.

Real packaging adds a further constraint: the pack must be **interlocking-free**, meaning every part can be removed without breaking or deforming the others. Parts hooked through each other like linked rings are not a valid pack, however dense.

## Geometry, solved in the frequency domain.

1. **Voxelize.** Parts and the container are represented as voxel grids, the 3D equivalent of pixels.
2. **Score every position at once.** Collisions and gaps between a part and everything already placed are computed as correlations, using the fast Fourier transform (FFT). One transform scores every candidate position of the part simultaneously.
3. **Place densely.** A cost function, also computed with FFTs, favors positions that minimize gaps against the container walls and neighboring parts.
4. **Stay interlocking-free.** Interlocking detection and correction run in the same framework, so the finished pack can be disassembled.

## Dense at scale, fast in practice.

- **35.77%[^2]**: Interlocking-free packing density, more than 6,000 objects from the Thingi10K set
- **40 seconds[^3]**: To place 670 objects at about 36% density
- **two hours at 37.30% density[^4]**: To arrange 6,596 objects

Benchmarks are from the published paper and MIT's coverage of it. Pack Studio applies the same engine to packaging, then ranks the result on landed cost.

## Where Pack Studio's engine came from.

**Pack Studio technology timeline**

| Date | Milestone | Source |
|---|---|---|
| July 2023 | MIT News reports the spectral packing method developed by MIT and Inkbit researchers | [MIT News](https://news.mit.edu/2023/chore-packing-just-got-faster-and-easier-0706) |
| August 2023 | "Dense, Interlocking-Free and Scalable Spectral Packing of Generic 3D Objects" published in ACM Transactions on Graphics and presented at SIGGRAPH 2023 | [doi:10.1145/3592126](https://doi.org/10.1145/3592126) |
| October 15, 2025 | Inkbit launches Pack Studio, bringing the packing algorithms from its additive manufacturing build preparation tools to packaging and supply chain engineers | [Press release](https://www.pack-studio.com/news/commercial-launch-of-pack-studio) |
| July 2026 | Pack Studio 1.10 adds Optimal orientation, rebuilt Stacking and Optimize Box | [Release notes](https://www.pack-studio.com/post/pack-studio-1-10-software-release-notes) |
| August to September 2026 | The Pack Studio Technical Series documents orientation, stacking and dunnage results | [Articles](https://www.pack-studio.com/blog/categories/technical) |

**Research record**

| Item | Detail |
|---|---|
| Paper | Dense, Interlocking-Free and Scalable Spectral Packing of Generic 3D Objects |
| Venue | ACM Transactions on Graphics, Volume 42, Issue 4, Article 141 (SIGGRAPH 2023) |
| Authors | Qiaodong Cui (Inkbit), Victor Rong (MIT), Desai Chen (Inkbit), Wojciech Matusik (MIT CSAIL and Inkbit) |
| DOI | [10.1145/3592126](https://doi.org/10.1145/3592126) |
| Open access copy | [MIT DSpace](https://dspace.mit.edu/handle/1721.1/152166) |

Pack Studio's lead technical inventor, [Qiaodong Cui, Ph.D.](https://www.pack-studio.com/authors/qiaodong-cui), is the paper's lead author and writes the [Pack Studio Technical Series](https://www.pack-studio.com/blog/categories/technical).

## Technology questions.

### What is Scalable Spectral Packing?

Scalable Spectral Packing (SSP) is the packing algorithm behind Pack Studio. It represents parts and containers as voxel grids and uses the fast Fourier transform (FFT) to score every candidate position at once, producing dense, interlocking-free packs. It was published in ACM Transactions on Graphics, SIGGRAPH 2023[^1]. See [Technology](https://www.pack-studio.com/technology).

### How is Pack Studio different from Pack Assistant?

Pack Studio runs on Inkbit's own packing engine, Scalable Spectral Packing (SSP), published in ACM Transactions on Graphics, SIGGRAPH 2023[^1]. Packaging and freight cost are built into every result, and each analysis ends in a one-click executive report. Lucid Motors describes the built-in cost estimates as something it has not seen in other packaging software. For a side-by-side comparison, bring a part to a demo.

### How fast are Pack Studio simulations?

Pack Studio goes from part CAD to box, pallet, container and truck layouts in under 90 seconds[^5]. At Lucid Motors a complete estimate, from CAD import to finished report with packaging and freight cost, takes 2 to 5 minutes[^6], down from 15 to 20 minutes[^7].

## See the engine on your parts.

Bring an awkward part. Watch it pack.

[Book a Demo](https://www.pack-studio.com/contact) · [How it works](https://www.pack-studio.com/how-it-works)

[^1]: Peer-reviewed publication: ACM Transactions on Graphics, SIGGRAPH 2023. Cui, Rong, Chen and Matusik; Inkbit and MIT CSAIL. Source: [Dense, Interlocking-Free and Scalable Spectral Packing of Generic 3D Objects](https://doi.org/10.1145/3592126), 2023-08-01.
[^2]: Packing density on a benchmark of more than 6,000 objects from Thingi10K, interlocking-free: 35.77%. Generic 3D objects packed into a cuboid container. Source: [Dense, Interlocking-Free and Scalable Spectral Packing of Generic 3D Objects](https://doi.org/10.1145/3592126), 2023-08-01.
[^3]: Time to place 670 objects: 40 seconds. About 36% packing density. Source: [The chore of packing just got faster and easier](https://news.mit.edu/2023/chore-packing-just-got-faster-and-easier-0706), 2023-07-06.
[^4]: Time to arrange 6,596 objects: two hours at 37.30% density. Large benchmark run. Source: [The chore of packing just got faster and easier](https://news.mit.edu/2023/chore-packing-just-got-faster-and-easier-0706), 2023-07-06.
[^5]: Time from part CAD to box, pallet, container and truck layout: under 90 seconds. Layout generation for one part across selected containers. Source: [One year in. Zero regrets. Lucid Motors one-year review of Pack Studio](https://www.pack-studio.com/lucid), 2026-06-23.
[^6]: Time per packaging estimate with Pack Studio, CAD import to finished report: 2 to 5 minutes. Start to report, including packaging and freight cost. Source: [One year in. Zero regrets. Lucid Motors one-year review of Pack Studio](https://www.pack-studio.com/lucid), 2026-06-23.
[^7]: Time per packaging estimate, historical workflow: 15 to 20 minutes. Carton density study, palletizing software and PowerPoint, before Pack Studio. Source: [One year in. Zero regrets. Lucid Motors one-year review of Pack Studio](https://www.pack-studio.com/lucid), 2026-06-23.
