Observable release boundary
Plain OCR V1 keeps document bytes and recognized text in the active browser tab. This page names the runtime requests, memory lifecycle, presentation dependencies, limits, and deferred capabilities you can verify.
Applies to the anonymous converter routes in Plain OCR V1
Selected, dropped, or pasted image bytes are validated and recognized in the browser. The OCR path does not call a route handler, server action, storage provider, database, billing service, or third-party OCR API.
Product path: File or Clipboard → browser validation → Web Worker → editable in-memory result
The Tesseract worker, WebAssembly core variants, and approved English and Simplified Chinese language data are served from versioned PicToTxt paths and load only after recognition starts.
Runtime prefix: /ocr/tesseract-v7/
The source preview uses a local object URL. Replacement and unmount revoke superseded URLs, cancellation terminates the worker, and refresh or tab close clears reachable workspace state.
No document job, result, usage, or retention row is created
Fontshare typography, fixed Picsum editorial images, and any configured site analytics or affiliate scripts can receive ordinary page requests. OCR code does not send them image bytes, recognized text, clipboard data, sensitive filenames, or object URLs.
External presentation dependencies are not OCR processors
In-memory lifecycle
One JPG, PNG, or WebP enters through picker, drop, or image paste.
Signature, declared type, decoded dimensions, and resource limits are checked.
A dedicated worker produces a canonical local result for review and export.
Replace, navigate away, refresh, close, or cancel to release local state and runtime.
Known limits
One JPG/JPEG, PNG, or WebP image. PDF, HEIC, TIFF, multi-page files, and batch input are not supported.
Maximum 10 MiB, 25 megapixels, and 8,192 pixels on either side. Browser memory and CPU still vary by device.
English or Simplified Chinese plus English. Handwriting and other languages are not implemented.
Focus, contrast, compression, layout, and typeface affect results. There is no accuracy or speed guarantee; review remains part of the product.
Not part of this release
The broader PicToTxt product direction includes structured and durable workflows, but none of the following is presented as implemented by Plain OCR V1.
Open the working tool, then use your browser network panel to observe the boundary.