Hi all, I wanted two things: a scanner I can use every day, where the document stays private and is easy to capture and manage, and a simple SDK so another app can scan a document straight in. Most SDKs on the market cost a lot for something that should be simple. Most scanners I tried wanted an account, then an upload to their cloud. The document was leaving the device just to become a PDF. After that I still had to email it to myself or find some other way to get it onto my PC. Scanicle does the whole job in the browser tab. A small model on your device (636 KB) finds the page in the live camera. Filters, quality checks, PDF building, and encryption stay on the device. No account, no upload, no analytics. After you see “Offline ready” (about a 29 MB download, mostly the signature checker), it works offline. The part that took the longest was Mobile Safari: getting it to run in the browser, stay inside Safari’s limits, and not crash. You can scan on your phone and manage the same document on your desktop. Pair two browsers with a QR code or a short typed code. This version does not stream the live camera to the desktop, and it does not let the desktop fire the shutter. You capture on the phone, then the finished page is sent over. Edits can happen on either device, but not on the same page at the same time. The original photo stays with each page, so either device can crop again. Pages go from device to device. The pairing server never sees a page. Import does not flatten PDFs into pictures. Text, fonts, and vectors stay. If you export a signed PDF as a whole file and do not change it, you get the same bytes. If you export only some pages or combine files, that becomes a new copy and the signature is gone. A local check against the EU trusted lists can mark a signature as verified only when the file already has the needed evidence and the bundled lists are still current. Otherwise it stays unverified until you allow an online check, one at a time. Limits, honestly: 50 pages per document, 50 MB per imported PDF, no OCR (an existing text layer is kept, but one is not created), and scans live in your browser’s storage — export the ones you want to keep. What still needs work: the SDK, the filters, a simpler and more polished UI, and encrypted browser storage. English and German. I would like to hear what you like, what should be better, other ideas worth adding, and what breaks on your device.

GPT-6 Astra Challenge

How did Astra change the scope or ambition of what you built?

Maker

GPT-6 Astra (Extra High) finally got this old idea running in the browser. I shelved it because it kept crashing, especially on iOS and older models couldn’t get it working. Astra fixed that. It’s 10–20× faster, and it actually works now. It also helped me retrain the edge-detection model. It even generated the edge-case images needed to fine tune it. This was my first time doing any machine learning.

2d ago

📌

3d ago

Scanicle

Scan documents to PDF in your browser, nothing uploaded

Scan documents to PDF in your browser, nothing uploaded