Voynich Viewer
Read the Voynich Manuscript (Beinecke MS 408) page by page, unfold its foldouts, and see the whole book as a block of
sheets in 3D: in its current binding, in the orders Lisa Fagin Davis has proposed, or in any order you make yourself.
Credit for the images belongs to Yale University (Beinecke Rare Book & Manuscript Library, MS 408). The research shown
here (collation, scribes, sections, the proposed orders) belongs to Lisa Fagin Davis. Voynich Viewer is an independent
project, not made or endorsed by either.
Site: voynichviewer.com.
- 3D (opens first): the book block in WebGL. Fan the sheets apart, pull one out, turn it over, unfold it, colour the edges by
scribe, section, illustration, language or quire, and see which faces touched when the book was closed.
Rearrange (key A) takes the book apart on the table: each quire lies open as a pile (sheets tucked inside each
other stacked with the centre sheet on top; sheets read one by one fanned like cards). It works like a drawing
program: click, ⇧/⌘-click or drag a box to select; drag onto a pile or between piles (a new quire); ⌘G new quire or
merge, ⇧⌘G split, ⌘] / ⌘[ toward the centre / outward, ⌫ set aside, ⌘X / ⌘V; a quire's name selects, renames
(double-click) or moves it; right-click menus. A panel in a fixed place shows the sheet pointed at (its place in its
pile, both its sides) or where dragged sheets will land. On a phone it opens as a panel under the book instead.
Hide lost sheets (key L) leaves the lost sheets out of the 3D book. A quire can be hidden in 3D, like a layer,
without changing the order.
- Reader: turn the pages two at a time, unfold the foldouts, zoom, jump to a folio. Zoomed in, a page is shown at
full size (about 3,600 px tall), loaded from Yale's IIIF image server. The Reader loads these ahead as you read (the
opening you are on, the next two and the one behind), so zooming in is sharp at once; not where zooming in gains
little, as on many phones, nor when the browser asks to save data. ▦ Grid (key O) shows every page at once,
quire by quire, with buttons for each section and for your bookmarks.
- Text in the Reader (key T): the text of the pages beside them, RF1b, René Zandbergen's reference
transliteration, shown as glyphs or written in Eva by his own tables. Point at a word on the
photograph to see what it reads; click it and the page stays put while the panel shows it cut upright from Yale's
photograph, with its other places as crops that open beside the page (N steps, Open page goes, Back returns). The
strip under the pages marks where the words are and where you have been. ⌘-click collects words to find the pages
that hold them in any order; "Pages that share these words" does it for a page; a ring's number unrolls the ring with
its repeats joined. Words round a circle are outlined as part of the ring, and the page turns to read them on request.
- Text tab (key /): search the whole book's text with a small query language (whole words,
* and ?, joins
within and across words, repeated words, regular expressions, filters such as scribe:2), narrow the results by
scribe, section, language, quire and kind of text, see the hits page by page in any order, open a result in the
Reader, export results as CSV or IVTFF, and save searches and page sets in Your work.
- Bookmarks: the ☆ in a page's corner (Reader), next to a page (3D), or key B; ★ at the top lists them. Name them,
and go to one: 3D and the Reader both follow.
- Info: credits, a plain-language guide, Davis's research (her 2025 post in brief, the Folio order tables built
like her quire diagrams for any order, her history of the book, the 2026 proposal), notes on single sheets, sources.
- Crop tool: re-cut any page from Yale's photograph (✂ Crop in the Reader, or in 3D). The photograph is loaded from
Yale's IIIF image server and cut in the browser.
The menu offers the current binding, "Davis/Layfield: proposed order" (every sheet read on its own, quires 13 and 20
re-ordered as in Layfield & Davis 2026, quire 9 re-sewn as in her 2025 post; the other quires keep today's sequence of
sheets, which is not part of the proposal), and your own orders. Old links to the
retired "Davis's blog (Jan 2025)" order open the Davis/Layfield order.
The 🐞 button (top right) shows how to report a bug or suggest a feature. The version (APP_VERSION in
assets/app.js, and the badge in index.html) is 1.0.
Your work is kept in the browser
There is no server and no account. Crops are kept in IndexedDB; your own orders, bookmarks, saved searches and page
sets in localStorage and, as a second copy, in IndexedDB. Your work (top right) exports all of it as a small progress file and imports it again, in this browser or
another one (the page images are cut again from Yale's photographs on import).
3D and the Reader follow the same place: the page you are on in the Reader is the sheet picked out in 3D, and the sheet
or opening you pick in 3D is where the Reader opens. Every view has its own URL (#read/<order>/<page>, #three/<order>/<view>), so a
view can be bookmarked or shared. The Reader's text panel adds /text (#read/beinecke/1r/text?w=f1r.15.8&n=2 for the
two words from the eighth of f1r.15, ?s=daiin+shol a set, ?like=f71r.L pages like 71r's labels, ?ring=f57v.3), and a search is #text/<order>/search?q=qok*&steps=scribe:2.
Run it locally
It is a static site: plain HTML, CSS and JavaScript, no build step. Serve the folder with any web server, e.g.
python3 -m http.server 8000
and open http://localhost:8000. (Opening index.html straight from disk does not work: browsers block fetch of
local files.)
To run the automatic checks (Node.js 20 or newer): npm install, npx playwright install chromium once, then
npm test. GitHub runs the same checks on every pull request; see tests/README.md.
Layout
index.html the page
assets/app.js the Reader, Info (with the Folio order tables), the order logic and the shared place (POS)
assets/work.js your work: crops (cut in the browser), your own orders, export and import
assets/arrange.js Rearrange in 3D: the table's selection, bar, menus and keys (the table itself is drawn in
view3d.js), the list beside it, and the panel under the book on phones
assets/view3d.js the 3D view (ES module, loaded when the 3D tab first opens)
assets/text.js the Reader's text panel, the word card and the interlinear (ES module, loaded when the text opens)
assets/search.js the Text tab: search, facets, the book strip, export (ES module)
assets/query.js the search query language: parse, say back in words, compile to a regular expression
assets/search-worker.js runs the searches off the page's thread
assets/fonts/voynich-vv.woff2 Voynich VV: Glen Claston's public-domain v101 font, with rare glyphs and kerning added
assets/privacy.js the cookie strip and consent; loads Microsoft Clarity only after "Accept"
assets/style.css
assets/vendor/ three.js r184 (MIT)
data/codex.json the physical model: sheets, quires, panels, scribes, sections, page variables, and for each
panel its Yale photograph (IIIF id) and the corners it is cut from; `mask` (f89v2) is
the part of the photograph blacked out, another page showing past a torn edge
data/orders.json reading orders, sources and per-sheet evidence notes; edited by hand
data/changelog.json what changed and when, newest first; shown under the version number (top left)
data/panels/ two JPEGs per panel face: _l (1400 px tall) and _s (300 px); zoomed in, the Reader loads the
page at full size from Yale instead
data/seams.json where the paper starts on a foldout's hinge panel: its photograph shows the stacked edges of the
book on the side the flaps hang, and the unfolded Reader lets the flap overlap them
data/text/rf/ the Reader's text (built by tools/text/rf.py): text.json (RF1b in STA, every line), alpha.json
(what each glyph is written as in each alphabet), shapes/<page>.json (where each word is on the
photographs, and its shape: a box, a turned box, or a band of a ring)
data/text/ Search's data (built by tools/text/build.py): pages/<page>.json (every line's consensus, its
split votes and gaps, and each transcriber's own line), index.json (every line's consensus, for
search), w/<code>.json (one transcriber's lines), meta.json (credits, method, page list),
glyphs.json (which font character draws each glyph)
tools/text/ the text pipeline: fetch.py (the pinned transcriptions), build.py, boxes.py, rf.py, rosettes.py, report.py,
test_text.py; the method is docs/TEXT.md section 3, the Reader's text docs/text/REDESIGN.md 8
tools/font/ builds Voynich VV from Claston's font (build_font.py), with test_font.py
docs/TEXT.md the design of Text: sources, method, data, query language, screens, tests
data/folds.json which way each foldout's folds go and which faces touch when the book is closed, as Lisa
Fagin Davis described them; 3D folds the sheets by it and marks those contacts confirmed
tools/import_from_scout.py refreshes data/codex.json and data/panels/ from a local Voynich Scout checkout
tools/seams.py measures those bands from data/panels/ and rewrites data/seams.json (run it after an import)
CNAME the custom domain for GitHub Pages
tests/ the automatic checks (Playwright); what they cover and how to read a failure: tests/README.md
.github/workflows/tests.yml runs them on GitHub for every pull request and every push to main
What's new: logging changes and fixes
Every change people can see, and every bug fix, gets an entry in data/changelog.json, in the same commit as the
change. Releases are newest first, each with its version, the date it went live (YYYY-MM-DD) and its changes,
each with a kind (new, improved or fixed) and a short line written for readers, not developers. Keep the list
short: one line per change, a few words each, only what a visitor would notice; a release of fixes is usually one
line. The site reads
this file: the badge at the top left shows the latest version, clicking it opens the list, and someone who has been
here before sees a dot on the badge until they have opened the newest notes.
Versions: the third number goes up with fixes (1.5.1), the second with new features (1.6), the first with a redesign.
Change APP_VERSION in assets/app.js and the fallback in index.html (#cx-ver) at the same time.
Credit people who report bugs by name only if they have said that is all right; otherwise "Reported by a reader".
Editing the orders
data/orders.json holds the reading orders. Each order lists its gatherings front to back; a gathering is nested
(a normal quire, bifolia outer to inner) or singulions (each bifolium read on its own). An order with base copies
another order and swaps in the gatherings in replace. Per-sheet options: spine (sew at another fold) and
inside_out. evidence holds the notes shown for a sheet; each src must be a key of sources.
Refreshing the page images
The panel crops and the physical model are made in Voynich Scout (scout codex-build and its crop editor). To copy a
new build here:
python3 tools/import_from_scout.py --dry-run
python3 tools/import_from_scout.py
Hosting on GitHub Pages
- Push this repository to GitHub.
- In the repository's Settings → Pages, deploy from the
main branch, root folder.
- The
CNAME file already names voynichviewer.com. At your domain registrar, point the apex domain at GitHub
Pages with A records to 185.199.108.153, 185.199.109.153, 185.199.110.153 and 185.199.111.153, and add a
CNAME record for www pointing to <your-username>.github.io. Then tick Enforce HTTPS once it is offered.
The .nojekyll file makes Pages serve the files as they are.
Analytics, cookies and privacy
Microsoft Clarity (visitor analytics, project ys2r2nospw) is loaded by assets/privacy.js, and only after a visitor
clicks Accept in the thin strip at the bottom of the page. Reject, or closing the strip, means no: Clarity is
never loaded. A Global Privacy Control signal also counts as no. The choice is kept for 6 months and can be changed in
Info > Privacy and cookies, which is the site's privacy notice. Clarity never runs on localhost or 127.0.0.1.
A separate, cookieless visit counter (Cloudflare Web Analytics) runs whatever the visitor chose, because it sets no
cookies and keeps nothing in the browser; Reject only turns off Clarity. It is the one request made before a choice.
It is off with a Global Privacy Control signal, on a local copy, and while CF_TOKEN in assets/privacy.js is empty. The
token is the site's Cloudflare Web Analytics token (Cloudflare dashboard > Analytics & Logs > Web Analytics > Manage site;
the token inside data-cf-beacon). It is public, so it is fine in the repository. The privacy notice has its "visit
counter" section only while the token is set, so to switch the counter off, empty CF_TOKEN.
CONTACT in assets/privacy.js is the address the notice gives for questions about data; change HOST there if the
site is not hosted on GitHub Pages. In Clarity's project settings
you can also turn on "Cookie consent", so that Clarity itself waits for the consent signal the site sends.
Credits
- Photographs: Beinecke Rare Book and Manuscript Library, Yale University, MS 408, from
Yale University Library's digital collections.
- Collation, scribes and sections: Lisa Fagin Davis,
"Voynich Codicology" (2025).
- Foldout structure: René Zandbergen, voynich.nu. Illustration type and Currier language:
the page variables of his ZL transliteration.
- Text: transcriptions by René Zandbergen and Gabriel Landini, Glen Claston, Takeshi Takahashi, the First Study Group
(William Friedman), Prescott Currier and Mary D'Imperio, Jorge Stolfi, John Grove, John Tiltman, Don Latham, Karl
Kluge (from Theodore Petersen's copy), Mike Roe and Denis Mardle, from René Zandbergen's
voynich.nu (CC0) and the Landini–Stolfi interlinear.
- Glyph font: Glen Claston's Voynich font (2005, public domain; UTF-8 fix by William Porquet), as Voynich VV.
- 3D: three.js (MIT licence,
assets/vendor/three.LICENSE).
