[mberman84]

Problem

Folio can recognize English in a scanned PDF. A reader opening a scan in another language, or a document that mixes languages, has no way to tell the reader which models to use.

Change

Settings → Reading has a searchable Scan languages dropdown with 123 language and script options. English ships with Folio. Readers can select additional languages, see the total download size, and save the packs for offline use. Applying the selection re-reads open scans and remembers the choice for future documents. Each download is limited to the expected size and checked against its pinned SHA-256 digest before Folio saves it.

Issue or approval

Approved scope for #15. This implements language selection in Folio’s built-in OCR. #21 covers document-wide search and searchable copies.

How I checked it

• On Windows, TypeScript, the production build, and the full Electron smoke suite passed. The regression rejects a damaged Spanish pack, retries successfully, recognizes a mixed Spanish and German scan, and reuses the saved packs.
• Downloaded French on Windows and verified it remained available after a full app restart with downloads blocked. In the browser edition, downloaded Spanish and Japanese, reopened the scan with external requests blocked, and selected the recognized text.
• Initialized all 123 model packs locally. The PR workflow runs the build and smoke suite on Windows, macOS, and Linux.

Before

After

Silent walkthrough

walkthrough.mp4

Implemented and verified with GPT-6 in Codex Desktop.