A scanned catalog arrives as one TIFF file. Your image endpoint returns a crisp, fast-loading picture. The integration looks finished—until someone asks where the remaining pages went. Cloudflare’s November 17 announcement makes that question timely: Images now accepts TIFF input, but decodes only the first page of a multi-page file. That is useful for a cover preview. A complete document viewer needs another path.
AI disclosure: This article was generated with AI assistance using Keith Azodeh’s public project documentation and the linked primary sources. It has not been reviewed by Keith before publication.
The launch removes a conversion step for teams serving an image from a TIFF master. Cloudflare says it can produce resized versions and formats such as WebP or AVIF without modifying the original. The source can remain intact while the delivered representation shows only part of it. That difference belongs in the application’s data model and interface.
One file is not always one picture
Keith’s Brain, Reconstructed case study offers a concrete example of why the contents of an image collection matter. For the November 2026 browser rebuild, he returned to DICOM files, separated the imaging series, and ordered slices by physical position. Sequential filenames alone were insufficient. The result uses a volume drawn from 30 measured slices and provides three movable cutting planes.
That project uses DICOM, Python, Three.js, and WebGL. The case study does not describe a TIFF pipeline or Cloudflare Images integration. Its relevant engineering lesson is the work required to turn a collection into the intended view: identify what belongs together, preserve its order, then choose how someone explores it. This is a personal visualization project, with no clinical validation claimed.
For a proposed multi-page TIFF workflow, the comparable question is simpler: does the consumer need the first page, one selected page, or every page? Those are different delivery requirements, even when the upload field accepts the same file extension.
A catalog card might need only a thumbnail. A document review screen needs navigation through all its pages. An archival download needs access to the retained source. Give each of those outputs an explicit purpose.
Make the preview’s scope visible
Consider an illustrative asset record for a twelve-page product catalog. Alongside its source identifier and version, store an independently inspected page count. Describe the image derivative as a first-page preview and record the source page it represents.
The interface can then say “Preview: page 1 of 12” and offer the full document through the application’s authorized delivery route. A generic “View image” label leaves readers guessing whether they have seen the whole asset.
For a full viewer, a possible design is to extract pages with a suitable processing tool, assign each page a stable identifier within the source version, and produce separate web derivatives. This is a proposed architecture, not a capability demonstrated by the new Cloudflare announcement. Its costs include extraction work, additional stored objects, and invalidating derivatives when the source changes.
Keep the original linked to that manifest. Otherwise, a corrected upload can leave page thumbnails associated with an earlier edition, or a reviewer can download a different source from the one that produced the preview.
Test the page that is easiest to miss
A successful HTTP response and a plausible first page are weak evidence for a multi-page viewing requirement. They can both pass while the feature is incomplete.
A small acceptance fixture can expose the difference. Create a non-sensitive three-page TIFF with large, unmistakable labels: COVER, SPECIFICATIONS, and ORDERING. Use distinct colors so a reviewer can recognize each page without reading fine print.
Check the proposed workflow against its actual promise:
- A preview-only route returns COVER and labels the result as the first page.
- A full-document route makes all three pages accessible in the expected order.
- Replacing the source updates the page count and derivative associations.
- Downloading the original returns the source version associated with the displayed preview.
These are suggested acceptance checks; they have not been executed against Cloudflare for this article. Their value is that they test what a user can reach, rather than treating successful decoding as evidence of completeness.
Budget for the source you actually have
Large masters introduce a separate constraint. Cloudflare’s published limits distinguish remote transformations from hosted uploads: the listed file-size limits are 100 MB and 10 MB respectively, while the Images binding has a 20 MB input limit. Verify the route your application uses before accepting a large TIFF into a workflow that depends on its transformation.
Cloudflare’s format table also explicitly lists TIFF with first-page support. Read format support alongside the page coverage your feature promises. Check the documented limits whenever you change delivery routes; accepting the same file type does not mean every route has the same capacity.
For publishers, commerce teams, and collection managers, today’s launch can simplify preview delivery. The practical decision is what the preview is allowed to stand for. Keep that promise small and visible: one delivered picture, clearly tied to the source and the pages it represents.
Top comments (0)