AI Hyperspectral Sorting: How Textile Recycling Reached Industrial Scale

Ask anyone working in textile recycling what the hardest part of the process is, and the answer is almost always the same: knowing what you have.

A bale of post-consumer clothing collected from a municipal take-back program contains polyester, cotton, nylon, blended fabrics, elastane, wool, coatings, finishes, and combinations of all of the above. Before any of it can be recycled, it needs to be sorted by fiber type. Get the sorting wrong and downstream processes fail: enzymatic depolymerization requires high-purity polyester feedstock; mechanical recycling produces lower-quality output if fiber types are mixed; chemical processes contaminate if the wrong materials enter the stream.

For decades, textile sorting has been done by hand. Trained sorters identify fiber types by feel, label, and experience, separating garments into categories at rates that are accurate but slow. At typical manual throughputs, sorting becomes the constraint that prevents recycling from reaching industrial scale.

The Scale Problem

Manual sorting can process roughly 0.3 to 0.5 tons per hour per worker under good conditions. At that rate, achieving 100,000 tons of sorted annual feedstock would require several hundred dedicated sorters working full shifts, year-round. The economics are prohibitive, and the consistency is inherently limited by human variability.

The textile recycling industry needed automation. But automating fiber identification is not a simple computer vision problem. Fabrics look similar across fiber types. A white polyester t-shirt and a white cotton t-shirt are visually indistinguishable to a camera. Fiber content has to be identified at the molecular level, not the visual level.

How Hyperspectral Imaging Works

Hyperspectral imaging captures light across hundreds of wavelengths simultaneously, well beyond the visible spectrum. Different polymer chains and fiber structures absorb and reflect light at characteristic wavelengths, producing a spectral signature unique to each material.

Where a standard camera sees color, a hyperspectral sensor sees molecular composition. Polyester, cotton, nylon, wool, and blended fabrics each produce distinct spectral fingerprints. Combined with machine learning models trained on thousands of garment samples, the system can identify fiber type, estimate blend ratios, detect contamination, and sort accordingly, in real time, at industrial throughput.

FastSort-Textile: 2+ Tons Per Hour

DataBeyond's FastSort-Textile system, deployed at Weavive's Shanhesheng facility in Zhangjiagang, applies this approach at a throughput of 2+ tons per hour. TIME magazine named it one of the Best Inventions of 2025, recognizing both the technical achievement and its potential impact on fashion's waste problem.

At 2+ tons per hour, a single sorting line can process volumes that previously required entire sorting facilities staffed by dozens of workers. The accuracy is consistent regardless of shift, volume, or fatigue. And the system produces structured output data on fiber composition and contamination that feeds directly into downstream process optimization.

For Weavive's ecosystem, the AI sorting line does more than sort. It is the intelligence layer of the entire chain. Sorted polyester flows to enzymatic depolymerization, where purity is critical to process efficiency. Sorted cotton, wool, and cashmere flow to mechanical recycling. Nylon is identified and routed to a separate stream for BASF-partnered recycling. Mixed-fiber residuals that cannot be cleanly sorted are mechanically processed into textile popcorn and powder as enzymatic feedstock, or into recycled cardboard.

Nothing is wasted. And that zero-waste outcome is only possible because the sorting upstream is accurate.

Beyond Economics: Traceability at the Point of Sort

Sorting is also where material traceability begins. Weavive uses Aware™ tracer technology, embedding physical markers in feedstock at intake, so that every batch of sorted material carries an identity that persists through depolymerization, repolymerization, and extrusion.

When a fabric mill or brand receives GRS-certified recycled polyester filament, the Aware™ data allows them to verify the provenance of the material, which collection region it came from, which processing batch it passed through, and what its fiber composition was at intake. This documentation is what the EU's Digital Product Passport requirements, taking effect from 2027, will demand.

The AI sorting line is not just a throughput solution. It is the starting point for a traceable, verifiable recycled content claim.

What Changes When Sorting Is Solved

Solving textile sorting at scale changes the economics of the entire recycling chain. Enzymatic depolymerization requires consistent, high-purity polyester input. Mechanical recycling quality improves with fiber type consistency. Polymerization produces better output when contamination is controlled upstream.

The sorting step is not one stage among many. It is the condition on which everything else depends. Getting it right, at scale, with machine consistency, is the infrastructure investment that makes post-consumer textile recycling commercially viable.

It is also why TIME called it one of the most important inventions of 2025. Not because it is technically impressive, though it is. But because it removes the constraint that has kept textile recycling from reaching its potential for decades.