YarnScope
Issue № 001Spring / 2026Klaipėda · A stash tracker for serious crafters
01The Tool · Ball-Band Scan

Scan a yarn ball band. We read the rest.

Point your phone at any ball band — Drops, Madelinetosh, Knit Picks, Cascade. YarnScope reads the brand, fibre, weight, yardage, and dye lot in under five seconds. No typing, no spreadsheet, no third-party site.

What a ball band actually carries

A standard ball band is a compact data sheet wrapped around the skein. From a single 4×8cm strip, a knitter learns: brand and yarn line; fibre composition and percentages; weight category (lace, fingering, DK, worsted, bulky); yardage and metres per ball; ball weight in grams; recommended needle and hook sizes; gauge over 10cm; care instructions; dye lot number and colourway.

That's a lot of information to type into a spreadsheet at 11pm.

How YarnScope reads it

You open the app, tap Scan, and frame the band in the viewfinder. The phone's camera captures a single still photo. Our OCR engine — fine-tuned on thousands of ball bands from the brands below — extracts each field in turn: brand first, then fibre, then weight, yardage, dye lot, recommended needles. The whole pass takes under five seconds on a modern phone.

You see the parsed values in a single review card. Tap to correct anything that misread; save to your stash. The next time you reach for that yarn, every field is searchable — and the colour swatch shows up in the right weight category, in the right project shelf.

Brands the OCR is trained on

The database is opinionated and growing. As of autumn 2026:

  • Drops Design and Garnstudio — multilingual, the OCR standard
  • Malabrigo — Rios, Worsted, Sock, Mecha, Lace, Twist, Mechita
  • Madelinetosh — Tosh DK, Tosh Sock, Tosh Merino Light, Vintage
  • Knit Picks / WeCrochet — the full house line
  • Cascade Yarns — 220, 220 Superwash, Heritage, Eco
  • Brooklyn Tweed — Shelter, Loft, Arbor, Vale
  • Quince & Co. — Lark, Chickadee, Tern, Owl, Piper
  • Rauma, Sandnes Garn, Isager, Holst Garn — the Nordic core
  • We Are Knitters — The Wool, Petite Wool, Pima Cotton
  • Wool and the Gang — Big Love, Crazy Sexy Wool, Shiny Happy Cotton
  • Plus several dozen indie dyers from the Ravelry forum corpus

If your brand isn't here, one corrected field teaches the engine — and the next scan from anyone reads it.

When the OCR misreads (it sometimes does)

Real ball bands curl on a skein, photos happen under kitchen lighting at 22:00, and one of the dye-lot digits gets smudged by a paper-shop sticker. YarnScope expects this. Misreads appear as orange-highlighted fields on the review card. Tap to correct. The corrected value is what saves to your stash — and an anonymised correction is sent back to improve the engine for everyone.

Privacy: what we keep, and what we don't

The photo of the ball band is processed transiently. By default, the image is not saved to our servers — only the extracted text data is associated with your stash entry. If you choose to attach the photo to the yarn (useful as a visual cue for thick-yarn-bag organisers), it stays in your account, encrypted at rest, on Supabase EU (Frankfurt). Delete attached photos or your whole account at any time — start@djump.io.

Questions about the scan

Does the OCR work in low light?
Yes if the band is in focus. Auto-exposure handles most kitchen lighting. If the photo is blurry, the review card prompts a retake before save.
What if my yarn brand isn't in the database?
Scan it, correct any misread fields, save. The corrected values teach the OCR engine, and the next person scanning that brand benefits. Email start@djump.io to prioritise a brand.
Does YarnScope read non-Latin ball bands (Cyrillic, Japanese)?
The Latin-script database is robust. CJK is on the autumn-2026 roadmap, prioritising Daruma, Hamanaka, and Olympus.
Is OCR available on the free plan?
OCR is a Pro feature at $3.99/month. The free plan supports manual entry of up to 50 yarns.
How long does each scan take?
Three to five seconds end-to-end on an iPhone 12 or newer; five to eight on mid-range Android. The bottleneck is camera focus, not OCR inference.