Imaging & IIIF

Manuscript scholarship depends on the image. Archetype treats the page image as a first-class, standards-based research object: every manuscript surface is delivered as a deep-zoomable IIIF image, described by a portable IIIF manifest, and — where an editor has linked transcription to the page — searchable region by region. Because the platform speaks the International Image Interoperability Framework (IIIF), the images and their descriptions are not locked inside one website. Any conformant viewer, and any other IIIF-aware repository, can consume them.

Deep-zoom images (IIIF Image API 3.0)

Page images are served by an embedded IIIF image server (SIPI), which exposes each scan through the IIIF Image API 3.0. A reader can pan and zoom to the level of individual pen strokes without the platform ever shipping a full-resolution file: the viewer requests only the tiles and resolutions it currently needs. In the frontend, OpenSeadragon reads each image’s info.json and renders a tiled pyramid; where a server advertises no usable tiles (none at all, or scale factors that are not powers of two, which SIPI can emit), the client synthesises a resolution pyramid from the declared sizes so zoom still works.

This design has practical consequences for institutions. Bandwidth scales with what a reader actually looks at, arbitrary crops and sizes are generated on demand (region/size/rotation/quality), and derivative images for thumbnails, comparison strips, or figures in a publication are all just IIIF URLs. The image server runs behind a reverse proxy restricted to read-only image delivery; its control and upload routes are never exposed publicly.

Adding and moving images

Curators add images from the backoffice manuscript workspace: files (TIFF, JPEG, PNG or JPEG 2000) are uploaded in resumable chunks, and a background worker converts each to lossless JPEG 2000, checks that the image server can actually render a tile from it, and only then registers the image. An image can also be moved to another item part from its edit dialog; a saved link that still carries the old part redirects to the image’s new address.

Portable descriptions (IIIF Presentation 3.0)

Each manuscript part is described by a IIIF Presentation 3.0 Manifest, generated on demand at /api/v1/iiif/item-parts/{id}/manifest. The manifest lays out one Canvas per image, ordered by locus, each carrying a painting annotation that points at the full image delivered by the IIIF image server. (The manifest deliberately omits the tile service for now, because SIPI’s non-power-of-two scale factors break tiling in common viewers; Archetype’s own viewer reads the image service directly.) Drop that manifest URL into Mirador, the Universal Viewer, or a colleague’s own IIIF workspace and the manuscript opens — no account, no export, no bespoke integration.

Where an image carries an editorial transcription or translation, the manifest also attaches a transcription AnnotationPage, and the image’s glyph annotations follow in a second page. These follow the W3C Web Annotation model: each annotation uses the supplementing motivation, carries the transcribed phrase as a TextualBody, and targets a specific rectangle of the page via a #xywh= media fragment. The regions come from transcription text that an editor has explicitly linked to an image region (a TEXT-typed annotation), resolved through the TEI references in the transcription. Canvas dimensions are read from each image’s info.json (cached, resolved concurrently, with a safe fallback), and legacy bottom-left coordinates are flipped into IIIF’s top-left origin so every box lands where it should. See Annotation & transcription for how those regions are authored, and the Data model for how images, texts, and graphs relate.

Comparing manuscripts

Because every part has a manifest, comparison is built on it. From a manuscript page or a manuscripts search, a reader can stage up to four manuscripts and open them side by side in an embedded Mirador viewer on the Compare page.

Search within the page (IIIF Content Search 2.0)

The manifest advertises a SearchService2 endpoint, so a IIIF viewer’s “search this manuscript” box works out of the box. A query returns an AnnotationPage of highlighting annotations, each boxing the region on the page whose linked transcription phrase matches — the reader sees the hit drawn directly on the parchment.

This is deliberately scoped, and worth stating plainly: matches are limited to transcription phrases an editor has manually linked to an image region. There are no per-word coordinates for unlinked prose, because the platform does not currently store HTR/ALTO word-level geometry, so an arbitrary word outside a linked phrase has no region to return. Full-text discovery across the corpus is handled separately by the search index — see Search & browse.

All IIIF endpoints are public and read-only, and honour the same visibility rules as the rest of the platform: an anonymous request sees only published transcription content. This makes a manuscript’s manifest safe to share and cite while editorial work in progress stays private.

Because these are open standards end to end — IIIF for images and structure, W3C Web Annotations for the region-anchored text — an Archetype instance is a good citizen of the wider manuscript ecosystem rather than a silo. See Interoperability for the full picture.