Listing your first item: what changed this cycle on Tutus

In March, the average first-time seller on Tutus spent four minutes and twelve seconds filling out form fields before publishing a single garment. Most of that time was lost deciding whether a jacket was "navy" or "midnight blue," hunting for fabric composition tags, and guessing a shipping package size that would not leave them out of pocket at the parcel shop.

This cycle, we stripped thirty-eight manual form fields down to four primary user actions. The objective was not to make listing mindless, but to transfer administrative labor from human typing to automated verification pipelines.

Automated Parsing Without the Hallucinations

When you drop three photos into the updated Tutus intake interface, the client-side pipeline converts raw camera files into a lightweight structural tensor. Instead of streaming uncompressed 12-megapixel photos over cellular data, the client extracts optical features locally and sends them directly to Mia's multi-modal intake service.

The service evaluates three primary signals: silhouette geometry, optical character recognition (OCR) targeted at wash tags and neck labels, and surface micro-texture. If you upload a knitted garment, the system distinguishes between heavy gauge cable knits and fine merino weaves in under 400 milliseconds.

Instead of forcing sellers to scroll through a drop-down list of 400 European brands, the interface presents a pre-filled suggestion with an explicit confidence rating. If the wash tag reads "100% Extra Fine Merino Wool - Made in Portugal," those attributes populate the product spec sheet automatically.

A Worked Example: Listing a Pair of Salomon XT-6s

To observe how this operates in practice, consider a seller in Lyon listing a pair of worn Salomon XT-6 sneakers.

Under the previous intake model, the seller opened Tutus, selected Footwear, selected Sneakers, typed "Salomon," typed "XT-6," selected size EU 43.5, selected color Ghost Gray/Silver, checked four distinct boxes for wear condition, estimated package weight at 1.2 kilograms, and searched third-party platforms to determine market value. Total elapsed time: 3 minutes, 40 seconds. Drops occurred frequently at field eleven when entering parcel dimensions.

Under the build shipped this cycle:

1. Photo Capture: The seller snaps three photos: lateral profile, tread sole, and interior size tag.
2. Attribute Extraction: Mia identifies the model variant (XT-6), extracts size EU 43.5 directly from the tongue tag text, and categorizes tread wear via edge-density analysis on lug depth.
3. Dynamic Valuation Range: The interface displays recent completed transactions for identical XT-6s across the Tutus network—showing an active band between €95 and €120—and defaults the listing price to the median €110 based on measured wear.
4. One-Tap Confirmation: The seller reviews four pre-populated fields, adjusts the price to €105 for a faster sale, and confirms.

Elapsed time: 38 seconds. Data completeness rose from a historical average of 64% to 98% in internal benchmarks.

The Underlying Engine: Trust Ledger and Escrow Routing

A low-friction listing process is counterproductive if it invites fraudulent items or inaccurate descriptions that fail at delivery. Tutus does not treat listings as isolated database rows; each published item interacts directly with WEVONE's core Trust Ledger and Escrow Hold Mechanism.

When a listing is submitted, Mia assigns a preliminary Risk Weight based on the seller’s unified Contribution Score across all WEVONE universes. A seller who has previously completed four successful short-term rentals in WEVONE Nest or fulfilled five verified tasks in Mission carries an established cryptographic trust history. For these accounts, the system bypasses manual moderation queues and enables instant listing visibility.

Simultaneously, the listing price locks into the Tutus Transactional Escrow framework. Buyer funds are held in a segregated European multi-sig vault until logistics tracking confirms delivery and a 48-hour inspection window elapses. If Mia detects a discrepancy between the seller's condition declaration and buyer-submitted unboxing media, the escrow automatically pauses payout and opens a structured dispute window where image tensors from listing day are compared pixel-for-pixel against incoming buyer claims.

Where the System Still Stumbles

Automated listing pipelines face distinct structural limitations, and we do not hide them behind polished release notes. During testing across 12,000 synthetic listings and 1,800 live beta uploads, three distinct edge cases caused failure modes:

• Unbranded Vintage: Garments from pre-digital eras lack standardized wash labels and modern sizing tags. When presented with a 1970s hand-knit cardigan, Mia’s OCR returns null and her geometry parser defaults to generic "outerwear." Sellers must manually input material content and sizing, bringing total listing time back up to approximately two minutes.
• Avant-Garde Silhouettes: Asymmetrical cuts, draped outerwear, and unstructured garments frequently confuse the optical texture engine, leading to misclassified garment categories.
• Fringe Postal Logistics: Localized shipping calculations in non-contiguous EU territories (such as the Canary Islands or French overseas departments) occasionally fail to auto-assign correct parcel weight tiers, requiring manual carrier overrides.

Retraining the garment-segmentation model on non-standard cuts and adding regional carrier fallbacks remain open technical tickets for the upcoming cycle.

What Shipped vs. What Is Planned

To distinguish present functionality from future product direction, here is the status of the Tutus intake architecture:

• Shipped (Fact): Client-side tensor extraction, multi-modal label OCR, automated attribute pre-fill, historical price band matching based on completed Tutus transactions, and Contribution Score bypass for instant listing.
• In Beta (Testing): Tread-wear auto-grading for technical footwear, material composition confidence scoring from close-up fabric micro-texture photos.
• Planned (Ambition): Cross-universe inventory routing allowing users to list an item for rent on Nest and for sale on Tutus simultaneously under a shared escrow lock.

Lowering the barrier to publish an item should never compromise catalog integrity. By replacing manual form entry with deterministic multi-modal extraction and anchoring listings to WEVONE's cross-universe trust architecture, this release treats cataloging as a programmatic background task rather than data-entry work.