FileNest's two OCR modes have different data flows
Local Privacy mode uses Tesseract.js in the current browser; selected images are not uploaded for OCR recognition. OCR runtime assets or language data may download on first use. High Accuracy OCR follows the request route shown in the workspace before execution. Only requests admitted by the current release status, browser, language and device conditions use a FileNest browser Worker without sending source images to an external OCR API; other requests use the disclosed transient Azure Document Intelligence prebuilt-read cloud route. A FileNest browser request that has started does not automatically switch to external OCR after failure. This conditional routing description does not mean FileNest-owned OCR has fully replaced Azure.
Start with the sensitivity of the source. If it contains identity material, an unpublished contract or private communications and local recognition is already sufficient, privacy-first processing may be the better tradeoff.
Keep the current production provider separate from next-generation FileNest OCR R&D
High Accuracy OCR follows the request route shown in the workspace before execution. Only requests admitted by the current release status, browser, language and device conditions use a FileNest browser Worker without sending source images to an external OCR API; other requests use the disclosed transient Azure Document Intelligence prebuilt-read cloud route. A FileNest browser request that has started does not automatically switch to external OCR after failure. This conditional routing description does not mean FileNest-owned OCR has fully replaced Azure.
Accuracy is not determined only by local versus cloud
Sharpness, skew, glare, typeface, language and layout complexity can matter more than where the model runs. A clean screenshot may work well locally, while a heavily blurred image or complex table can still require review even after cloud recognition.
- Crop irrelevant background first.
- Correct skew and glare where possible.
- Review amounts, dates, names and account fields character by character.
- Do not treat complex-table OCR as final structured data without checking it.
When local fits and when cloud is worth considering
Local mode is attractive when privacy matters, the network is unreliable or the source image is already clear. If the device is constrained, the image is more difficult or the first local result is clearly inadequate, High Accuracy mode can be considered after checking the actual request route; only the cloud route sends source images to external OCR.
This is a selectable tradeoff rather than one mode universally replacing the other.
Human review still matters in both modes
OCR reduces manual typing; it does not remove verification. For legal, financial, identity or medical text, recognized output should be treated as text to verify rather than a replacement for the source or professional review.