I always thought the Treventus ScanRobot was an elegant way to scan books: [this video] It looks like it's sucking data right out of the book. Look for other videos of the process.
Elegant scanner that appears to suck data out of the book.
Generally: unbinding the leaves of a book so that they can be more easily handled by a scanner, often including an automated feeder. The simplest method is to simply cut off the binding (a powered paper cutter can do this in an instant). The alternative is nondestructive scanning, where the binding and publication as a whole remain intact. This is typically performed using flatbed scanners, angled scanners (the book is either face up or face down on a scanning bed which typically forms a ~90 degree angle), automated feeders which incorporate page turning (via numerous methods), or face-up scanning with digital deconvolution of page distortions. Higher speed presents higher risks of both poor / missed scans, and damaging original materials. Methods should be tuned to the materials, their intrinsic value and/or uniqueness, and end goals. See: [link] There are numerous book-scanning projects with YouTube videos of their methods. Typical rates range from ~300 -- 3,000 pages/hour, with 1,000 pp/hr being a good middling rate. DIY 1,000 pp/h [link] [link] Commercial, manual page turning: [link] 3,000 pp/h vacuum page turning [link] ... capable of managing very thick volumes: [this video] Google fully automated linear scanbot: [link] Face-up, no-platten, with deconvolution: [link]
There‘s a somewhat similar commercial scanner [1] [2], with a V-design as well but inverted to scan from the top. Much gentler on the books as it‘s the scanner that moves, not the books themselves. Super happy to see someone develop an open-source alternative! [1] [link] [2] [this video]
Commercial V-design scanner that moves the scanner, not the books.
Video list and Hacker News comment labels from HackerTube (prompted by @karpathy), released under ODC-By 1.0, built on the open-index/hacker-news archive. Hacker News data is provided by Y Combinator; comments remain the work of their authors. Video titles, channel names and thumbnails belong to their creators and are shown via YouTube. HackerTube Watch is an unofficial edition. Data export of 2026-10-31.
3 Hacker News comments
From the HN archive; up to 20 kept per video, favoring different commenters. Recommendations first.
Citations, criticism and jokes
Read more