Downloading a page from Google Books newspaper archive (high quality)

How to save a specific page of a Google Books-hosted newspaper (e.g. The Indian Express, 6 Jul 1992) as a PDF at the same quality the browser viewer shows. Works for volumes where can_download_pdf is false (no official PDF button) but the page is in full view.

Example used throughout: id=i39lAAAAIBAJ, page 13 (pg=PA13).

TL;DR

  1. Scrape the viewer HTML for the page’s sig token, tileres resolution tiers, and page_scanjob_coordinates.
  2. Quick path: append &w=2048 to the page image URL → single PNG capped at ~1918x2500 (slightly soft).
  3. Best path: drive the viewer with Playwright, sweep-scroll the whole page, collect the 256px tile positions from the DOM, canvas-stitch, screenshot → PDF (native ~2355x3072 or more).

1. Confirm the volume is full-view and identify the scanjob coordinates

UA="Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0 Safari/537.36"

curl -s -A "$UA" "https://books.google.co.in/books?id=<ID>&jscmd=SearchWithinVolume2&q=<anyword>"

Returns JSON with matching page ids — proves full-view access (no login needed for these archives).

2. Get the signed page URL + tile metadata

curl -s -A "$UA" "https://books.google.co.in/books?id=<ID>&pg=PA<NN>&source=gbs_selected_pages&hl=en" -o bookpage.html

Inside bookpage.html, find the JSON block for your page ({"pid":"PA13",...}). It contains:

Field Use
src page image URL incl. sig=... token (works from curl)
h, w full scan dimensions
additional_info.[NewspaperJSONPageInfo].tileres resolution tiers [{"h":1280,"w":981,"z":0},{"h":2048,...,"z":2},{"h":3072,...,"z":4},{"h":4352,...,"z":6}]
page_scanjob_coordinates {x,y} tile URLs use pg=x%2Cy, NOT pg=PA13
volume JSON [NewspaperJSONVolumeInfo].res_h the tier heights again

The same sig token works for every zoom/tile request of that page.

3. Quick method (single image, decent but capped)

curl -s -A "$UA" -e "https://books.google.co.in/books?id=<ID>&pg=PA<NN>" \
  "https://books.google.co.in/books/content?id=<ID>&pg=PA<NN>&img=1&zoom=3&hl=en&w=2048&sig=<SIG>" \
  -o page.png

sips -s format pdf page.png --out page.pdf

Notes:

4. Best method (tile stitching — true browser quality)

The browser renders the page from 256x256 tiles:

https://books.google.co.in/books/content?id=<ID>&pg=<X>%2C<Y>&img=1&zoom=<Z>&hl=en&sig=<SIG>&tid=<N>

Do NOT guess the tid→position mapping. It is grouped in 3x3 blocks with per-strip offsets and the viewer skips some edge columns — empirical guessing produced shuffled tiles. Instead let the viewer tell you the layout:

4a. Sweep-scroll the viewer and collect tile positions (Playwright)

Open https://books.google.co.in/books?id=<ID>&pg=PA<NN>&source=gbs_selected_pages&hl=en then run:

async () => {
  const sleep = ms => new Promise(r => setTimeout(r, ms));
  const scroller = document.querySelector('.overflow-scrolling');
  // Anchor to the ANCESTOR of a known tile img for YOUR scanjob coords — do NOT find the
  // page div by scrollWidth/Height equality: sibling pages in the scroll strip can have
  // same-sized divs, and width can be off by 1px from tileres (both caused silent 0-tile runs).
  const t0 = document.querySelector('img[src*="<X>%2C<Y>"][src*="tid="]');
  let pageDiv = t0.parentElement;
  while (pageDiv && !(pageDiv.scrollHeight >= 3070 && pageDiv.scrollHeight <= 3075 &&
        pageDiv.clientHeight === pageDiv.scrollHeight && pageDiv.clientWidth > 1500))
    pageDiv = pageDiv.parentElement;
  const sr = scroller.getBoundingClientRect(), pr = pageDiv.getBoundingClientRect();
  const pageTop = pr.top - sr.top + scroller.scrollTop;
  const pageLeft = pr.left - sr.left + scroller.scrollLeft;
  const map = {};
  const collect = () => {
    Array.from(pageDiv.querySelectorAll('img[src*="tid="]')).forEach(im => {
      const tid = +(im.src.match(/tid=(\d+)/) || [])[1];
      map[tid] = { l: parseFloat(im.style.left) || 0, t: parseFloat(im.style.top) || 0 };
    });
  };
  const maxL = scroller.scrollWidth - scroller.clientWidth;   // IMPORTANT: sweep ALL the way right
  for (let top = pageTop - 300; top < pageTop + 3072 + 300; top += 400) {
    for (let left = 0; left <= maxL; left += 400) {
      scroller.scrollTop = Math.max(0, Math.min(top, scroller.scrollHeight - scroller.clientHeight));
      scroller.scrollLeft = left;
      await sleep(650);
      collect();
    }
  }
  window.__tileMap = map;
  return Object.keys(map).length;
}

Gotchas that each cost a re-run:

4b. Stitch on a canvas (same page context, same-origin fetch works)

async () => {
  const map = window.__tileMap;
  const base = 'https://books.google.co.in/books/content?id=<ID>&pg=<X>%2C<Y>&img=1&zoom=<Z>&hl=en&sig=<SIG>&tid=';
  const canvas = document.createElement('canvas');
  canvas.id = 'stitched';
  canvas.width = 2355; canvas.height = 3072;            // tileres w x h for the tier
  canvas.style.cssText = 'position:fixed;left:0;top:0;z-index:2147483647;background:#fff';
  const ctx = canvas.getContext('2d');
  ctx.fillStyle = '#ffffff'; ctx.fillRect(0, 0, canvas.width, canvas.height);
  for (const [tid, v] of Object.entries(map)) {
    const r = await fetch(base + tid, { credentials: 'include' });
    const bmp = await createImageBitmap(await r.blob());
    ctx.drawImage(bmp, v.l, v.t);
  }
  document.body.appendChild(canvas);
  window.scrollTo(0, 0);
  return 'done';
}

4c. Screenshot the canvas → PNG → PDF

await page.locator('#stitched').screenshot({ path: 'page.png', scale: 'css', type: 'png' });

sips -s format pdf page.png --out page.pdf

A 2355x3072 canvas screenshots fine (CSS-pixel scale = exact device-independent capture).

5. Verify before shipping

Open the stitched PNG and check:

If you spot a hairline vertical seam (split/clipped glyphs, few-px vertical step): before debugging your stitch, check whether it’s in Google’s source scan — old pages were sometimes digitized in vertical strips. Fetch the server-side composite (quick method URL from §3) and look at the same spot; if the seam is there too, your stitch is faithful and nothing is fixable.

6. Misc facts