DocWeave: PDF Document Scanner

Turn PDFs, images, and paper documents into structured content you can actually edit and reuse.

DocWeave runs an AI document recognition model locally on your device. Import PDFs or images, scan paper documents, book pages, and lecture notes with your iPhone or iPad, or capture scrolling webpages and long articles through screen recording.

Once the local model is installed, recognition works without an internet connection, and your documents never need to be uploaded to a server.

What It Can Do

• Scan Paper Documents

Automatically detect and continuously capture pages on iPhone and iPad, then add them to the processing queue. Ideal for book pages, lecture notes, receipts, and temporary documents.

• Capture Scrolling Content

Record scrolling webpages or long articles and automatically stitch them into a long screenshot. Before recognition, you can remove ads, duplicate sections, and irrelevant content.

• Understand Complex Layouts

Go beyond plain text. DocWeave can recognize paragraph structures, tables, LaTeX formulas, code blocks, and algorithm blocks.

• Export in Four Formats

Generate reconstructed PDF, Markdown, EPUB, and MinerU packages in one pass—ready for reading, editing, archiving, or further processing.

• Preview as Recognition Progresses

View structured results as soon as each page is completed, so you can follow the progress and check the reconstructed layout.

• Resume After Interruptions

If a model download, page capture session, or large document recognition task is interrupted, continue from the saved progress instead of starting over.

• Keep Documents on Your Device

Model inference, page images, and recognition results are stored locally. Unless you choose to export, share, or send them to another app, your recognized content stays on your device.

From capturing paper and importing files to understanding complex layouts and exporting reusable content, DocWeave keeps the entire document workflow on your device.

First Use

An internet connection is required to download the local recognition model, which is approximately 2.5 GB. The download begins only after you confirm it and can resume if interrupted. Once installed, the model can perform recognition offline.

Device Requirements

On-device AI recognition requires an A14 / M1 chip or newer and at least 6 GB of RAM. Requirements differ by recognition engine:

• OvisOCR2 (faster, lower memory usage): a device with at least 6 GB RAM; keep at least 3 GB of free RAM before loading the model; model files are about 0.6–1 GB (varies by device).

• MinerU (full-featured, supports reconstructed PDF export): a device with at least 8 GB RAM; keep at least 5 GB of free RAM before loading the model; the model file is about 2.5 GB.

Note: free RAM is not storage; you also need enough storage for the model files and recognition results.

Recommended devices:

• iPhone 12 Pro / 13 Pro, iPhone 14 or newer (MinerU requires iPhone 15 Pro or newer)

• iPad mini (A17 Pro)

• iPad Air (M1) or newer

• iPad Pro (M1) or newer

• Mac with Apple silicon and at least 8 GB of unified memory

Devices before A14 are not supported, e.g. iPad Pro (2018/2020, A12X/A12Z chip).

Camera scanning and rolling-screen recording capture are only available on iOS/iPadOS. The macOS version supports importing PDFs and images via file picker, drag and drop, or paste.