Paper dolls and paper outfits, cut out again
Zelda Fitzgerald's paper dolls at the Beinecke Library: eight figures and six outfits, each one a separate piece of painted paper about 30 cm tall. Every piece here is its own draggable layer, pulled from Yale's IIIF Image service as a region of its own canvas, with the black photographic board it was shot on removed in the browser so the silhouette floats. Drag an outfit onto a figure and it stays exactly where you let go, painted above the figure; drop a second outfit on the same figure and it lands on top of the first.
The Sheet view shows the uncut arrangement. It is a reconstruction, not a photograph: this copy is already cut, so Yale's manifest has one canvas per piece and no sheet to quote. The sheet is derived from home positions recorded in one standalone Annotation Page, and Dress-up spreads the same pieces over a workspace. The page reads that annotation page together with Yale's own manifest, so the geometry it draws is assembled at load time rather than baked in; a derivative Manifest publishes the same model for viewers that cannot read a standalone annotation page, and it still points at Yale's image service. No image is copied here.
Assembling paper-doll model…
How the interaction works
Registration. A garment is never re-scaled to fit. All fourteen recto exposures come from one fixed camera photographing pieces of 28–31 cm, so a pixel of an outfit and a pixel of a figure are the same piece of the real world, and an outfit lands on a figure at 1:1. Where you let go is where it stays: the point under the pointer does not move.
Soft snap. Each figure carries one torso anchor, an upper-chest point recorded in the annotation page. If a garment's own anchor lands within 5% of the sheet width of a figure's, the garment eases the rest of the way so the two coincide. Turn the toggle off to see the difference: free placement still works, it is just less forgiving.
Stacking. There is no layer panel. Garments occupy a z band above every figure, and within that band each drop takes the next counter value, so the most recently dropped outfit is the one on top. Figures restack among themselves the same way.
Wearing. Once an outfit is on a figure it stays on it: dragging the figure takes its clothes with it, at the offset you left them. Dragging the outfit off detaches it again.
Transparency. Yale publishes JPEGs, which have no alpha, and the pieces are photographed on a near-black board rather than a white one. The page therefore floods inward from the border of each crop through pixels close to that border's median colour and marks everything it reaches as transparent, once per piece, and keeps the resulting bitmap. The tolerance was picked by checking the result against an independent segmentation of the same crops (mean intersection over union 0.989).
Grabbing. A cut-out piece is mostly transparent inside its bounding box, and a snapped outfit covers its figure almost completely, so a rectangle test would pick the garment when you aimed at the doll underneath it. The page walks the stack from the top down and takes the first layer with non-transparent pixels under the pointer.
Keyboard. In Dress-up mode a press focuses the piece it picked. Arrow keys nudge it by 0.4% of the stage, Shift by 2%, and each move finishes with the same drop logic a pointer gets, so the keyboard can snap an outfit to a torso and has the result announced, not just slide a layer around.
Yale catalog record · Yale IIIF Manifest · Paper-doll Annotation Page · Derivative Manifest
About this item
Original item: Paper dolls and paper outfits painted by Zelda Fitzgerald [graphic] , ca. 1927. Beinecke Rare Book & Manuscript Library, Yale University. Call number YCAL MSS 317; the digital object is 11873654.
The catalogue describes 8 paper dolls (5 male, 3 female) and 6 paper outfits (4 men's, 2 women's) in graphite, watercolor and gouache, 28–31 cm. It also records the condition of the thing itself: two of the outfits are taped onto the dolls they fit, the other four do not fit the dolls in this collection, three arms, two hands and two feet are missing from the dolls, one head was partly replaced by the artist, and three sleeves are missing from the outfits. The fourteen recto canvases in Yale's manifest are exactly that set, one per piece.
Yale's rights statement is the usual caveat that the material may be subject to copyright. This demo stores no copies: it fetches Yale's manifest and asks Yale's image server for regions at display size, and it attributes the object and links the catalogue record above. The reading of the object encoded in the annotation page (which piece is a figure and which an outfit, the sex each is cut for, and the torso anchors) is ours, and the two figures whose sex is not obvious on the scan were assigned so that the counts match the catalogue rather than from certainty about the drawing.
How this page is assembled
Two documents, and the numbers below are the ones the page actually derived at load time.