Reading across languages
From your own book to reading in another language
The challenge
A reader can have the right book in the wrong language. Globoox lets people bring an EPUB and read it in English, Spanish, French or Russian.
My goal was to make that journey straightforward, from adding a book to reading it in a chosen language. Readers should be able to change language, text size or device and keep their place.
My role
The journey needed to work across the website, library and reader. As the full-stack designer, I design the interfaces and shared UI, then implement them in the frontend. Working in the browser lets me check interaction and typography with real books.
Reading and translation need to work together. I work with the CTO and linguistics specialists on the experience around the translation engine: adding a book, navigating it, adjusting the page and returning to it.
See how it works before adding a book
People need to see what they will get before bringing their own file. I put real product recordings at the centre of the website and explained the first use in three steps.
An interactive excerpt lets visitors compare the original and translation. Editorial typography and botanical details give the website its character; quieter surfaces keep the app and reader focused on the book.
Get from EPUB to reading
Getting a book ready takes more than selecting a file. I made the EPUB requirements, preparation and processing visible, with specific recovery advice when something goes wrong.
Once signed in, add an EPUB, open the book and choose a reading language.
Add your book
Select an EPUB from your device.
Choose a language
Open the book, then select a reading language.
Start reading
Continue in your chosen language.
In the product
Keep translation inside the book
Changing language should not create another document to manage. The title, chapters and text follow the chosen language inside the same book. Pending text stays distinct from ready text, while nearby content is prepared as the reader moves on.
Keep the passage when the pages change
A reader needs the same passage after a change of language, text size or device. Saving a page number was not enough: the paragraph moves as the layout changes. Early restore approaches also produced clipped text and position drift.
We made the passage the stable reference: book, chapter and a position within the text. Page breaks are calculated around it. Matching the pagination measurements to the visible typography keeps lines from being clipped.
Typesetting for real books
Changing text size reshapes the whole page. I tuned line spacing, paragraph and list breaks, and language-aware hyphenation. Nonbreaking spaces keep short words with the words that follow. The paginator uses the same type settings as the visible page.
The first five tabs compare the same text with one rule switched off. Line and page-break examples use test content. Reader comparison shows an Apple Books screenshot.
Rule off
Rule on
Make room for different reading habits
The page should suit the person reading it. Text size, four themes and a single-page or spread layout give readers control without changing the library’s appearance.
The book stays in view until a control is needed. Shared drawers and dialogs keep closing, focus, scrolling and safe areas consistent across the interface.
Early product signals
We tracked whether new signups reached the reader and how quickly they uploaded a book. Observed in the beta, June–September 2026.
New signups who uploaded a book and opened the reader within 7 days
Median time to a successful upload, among converters
Reddit → signup
Tracked Reddit referrals who signed up within 7 days
How these were measured
PostHog production data from 16 June through 13 September 2026 UTC. Configured internal and test traffic was excluded. These are exploratory beta observations.
Activation: 22 of 41 tracked new signups completed signup → successful upload → reader ready, in order, within seven days. Reader ready means the interface was ready; it does not measure attentive reading. The median signup-to-upload interval among 22 converters was 131.97 seconds.
Reddit: 5 of 10 tracked people arriving through a Reddit referral on any production page signed up within seven days. This is a referral cohort, not the conversion rate of the landing page.
The period predates the current dev website release. Redesign uplift was not measured. Late cohorts may not have a full seven-day observation window; consent and client-side tracking affect coverage.