SHEET 08 — AGENT INTERFACE checking…

Archive QC pipeline

Clean, correctly-oriented scans — by the thousand.

InstaDocs batch-cleans scanned document archives — strips scanner borders, removes fold-shadows and ink-blots, fixes rotation, mirror and skew, then flags the suspects for review. 25 languages, 60,000+ pages a shift, entirely on your own machine.

PS> iex (irm https://get.appz.dev/instadocs/install.ps1)
restored
Raw court-document scan — black border strip and skew, content blurred for privacy RAW · border · skew
Cleaned output — borders stripped, page deskewed, content blurred for privacy CLEAN · 0.0°

← drag · raw scancleaned output →

60k+
pages / 8-hr shift
25
languages, script-aware
4
cardinal angles fixed
100%
offline — no phone-home

The problem

Archive QC is done by hand, one scan at a time. It doesn't have to be.

Without InstaDocs
  • Operators eyeball-rotate every scan by hand — slow, inconsistent, error-prone.
  • Scanner over-scan leaves black borders; fold-shadows and ink-blots bleed across pages.
  • Bulk tools flatten everything — wiping handwritten page numbers, seals and signatures.
  • No record of what changed: a bad batch means re-scanning thousands of files.
With InstaDocs
  • Orientation, mirror and skew detected automatically — operator confirms, never guesses.
  • Borders and patches removed while marginalia, thumbprints and seals are preserved.
  • Every run logged to a local history DB — fully auditable, byte-for-byte reversible.
  • Suspects surface to the top; clean pages pass straight through.

What runs on every page

Six deterministic steps. No black-box model.

Consistent, explainable steps — every page is handled the same way, and every decision is recorded so you can see exactly what changed.

  1. 01

    Strip black borders

    Scanner over-scan and paper-holder margins are removed — while handwritten page numbers and margin notes are detected and kept.

  2. 02

    Remove patches & streaks

    Fold-shadows, ink-blots and hairline scanner-glass streaks are cleared away, while thumbprints, signatures and seals are carefully preserved.

  3. 03

    Correct orientation

    Every page is turned the right way up — 0, 90, 180 or 270° — automatically, with extra care on the cases that trip up lesser tools.

  4. 04

    Catch mirroring

    Rare mirror-flipped scans are detected and flipped back the right way round.

  5. 05

    Deskew tilt

    Tilted pages are levelled, with safeguards that fix genuine skews and leave clean pages untouched.

  6. 06

    Emit clean output

    A cleaned image plus a per-file report — every decision recorded for the run history and review.

Automatic detections

It catches what slips past the eye.

Beyond cleaning each page, InstaDocs watches for the problems that quietly ruin an archive — and flags or fixes them before they reach the shelf.

Blank pages — removed, never dropped

Near-empty pages are spotted automatically and lifted out of the output, with the originals kept safe in backup. Optionally stamp a watermark on every blank, so nothing is ever silently discarded.

Duplicate pages — caught and reviewed

Near-duplicate scans within a batch are surfaced side by side. The operator decides — keep one, keep both, or drop — and every choice is logged with a reviewer and timestamp.

Blurred pages — flagged for a second look

Out-of-focus scans are flagged automatically, so a bad capture never slips into the archive unnoticed. Reprocess or rescan from the same flag workflow.

Faded pages — recovered

Low-contrast, faded pages are detected and brightened in place, bringing back text that would otherwise be lost — and helping duplicate detection match faded copies of the same page.

On real files

Pulled straight from a production run.

Actual pages from a court-document archive, before and after InstaDocs. Each is deliberately blurred — you can see the cleanup, not the case.

Over-scan border stripped — raw scan, content blurred for privacy BEFORE
Over-scan border stripped — cleaned output, content blurred for privacy AFTER

Over-scan border stripped

A solid black scanner margin whited out — the page content left untouched.

Edge border removed — raw scan, content blurred for privacy BEFORE
Edge border removed — cleaned output, content blurred for privacy AFTER

Edge border removed

Heavy black banding along the edge cleaned away, marginalia preserved.

Fold-shadows & blots removed — raw scan, content blurred for privacy BEFORE
Fold-shadows & blots removed — cleaned output, content blurred for privacy AFTER

Fold-shadows & blots removed

Patches masked out while thumbprints, signatures and seals are kept.

Real archive pages · content irreversibly blurred & metadata stripped for privacy

Script-aware OCR

25 languages, one pipeline.

InstaDocs detects the script on each page and selects the right language automatically — no per-batch flag to set. English ships in the installer; the other 24 download on demand from Settings → OCR Languages.

bundled install on-demand
InstaDocs OCR settings — language list with English default, Tamil installed, and on-demand install buttons for Arabic, Bengali, Chinese and more.
A English Tamil Hindi Bengali Telugu Kannada Malayalam Gujarati Punjabi Odia Assamese ا Urdu Marathi Nepali Sinhala Sanskrit ع Arabic پ Persian א Hebrew Chinese Chinese (Tr.) Japanese Korean Thai Vietnamese

Use cases & solutions

Built for the people who keep the record.

Wherever scanned documents are archived at scale, the same problems repeat. Here's how InstaDocs fits each one.

Courts & judicial archives

Decades of case files scanned at intake, rotated every which way.


Auto-correct orientation and skew at ingestion; flag suspect pages for a clerk to confirm before the file is sealed.

Land records & registries

Bulk deed and registry scans across many Indic scripts, heavy seals and stamps.


Borders and blots removed without erasing official seals or registration marks — the script is detected per page, so no language is mangled.

Libraries & heritage digitisation

Fragile manuscripts where a thumbprint or marginal note is part of the record.


Texture-preserving patch removal keeps signatures, thumbprints and handwritten annotations while cleaning the background.

Legal firms & e-discovery

Exhibit scans must be clean and upright before OCR and indexing.


Run a fast daily pass to straighten and de-border exhibits, raising downstream OCR accuracy with zero manual rotation.

Scanning bureaus & BPOs

High-throughput intake under SLA, tens of thousands of files per shift.


Daily mode clears 10k pages every ~33 min on 4 cores — 60,000+ in an 8-hour shift; CLI + Explorer integration drop it straight into existing batch workflows.

Government records depts

Air-gapped machines, strict audit and chain-of-custody requirements.


100% offline, per-machine licensed, every run logged with byte-verified backups — nothing leaves the building.

In the operator's hands

A toolkit, not a button.

Two run modes

Daily for fast production intake, thorough for periodic deep QA — one flag switches the whole pipeline.

Operator reviewer + editor

Whiteout, crop, brush and freehand-polygon tools with undo/redo, fully keyboard-driven for speed.

Suspect flag workflow

Problem pages surface to the top; flag bleedthrough, rotate, redo or suspect and reprocess just those.

Operator override, always

Every auto-decision has a visible override. The domain expert is the authority — not the model.

Byte-verified restore

In-place QC keeps a sha256-checked, byte-identical backup of every original. Revert a file or a whole run.

Full run history

Every run recorded to a local history database with trends — nothing scattered across loose files.

HTML review export

Generate a before/after thumbnail page for any run; open and share in any browser.

CLI + Explorer hooks

A full command-line tool and a Windows right-click menu drop QC into existing batch workflows.

Extract page text

Pull the recognised text from any page as plain text or a structured file, with an optional conservative clean-up.

Grid review

Triage a whole batch at a glance in a thumbnail grid — not just one page at a time.

Archive-ready export

Convert colour scans to compact black-and-white archive files in a single pass.

Annotate & mark up

Draw boxes, points and notes on a page that stay attached to the file and its history.

Configurable pipeline

Every step is optional. Configure what runs.

Six QC steps, six independent flags. Turn any of them off for a faster pass, or run the full set for maximum coverage — daily and thorough below are just two common starting points, not the only options.

Border strip --no-black-borders Skip scanner over-scan removal
Orientation detection --no-orientation Skip 0/90/180/270° auto-rotation
Skew correction --no-skew Skip tilt leveling
Mirror detection --no-detect-mirror Skip horizontal-flip check (faster)
Texture-preserving rule --no-keep-textured Disable thumbprint/seal protection
Signature-zone preserve --no-signature-zone Disable lower-right signature guard

Two starting presets

--mode daily
~33 min
10k files · 4 cores

Daily intake

Patches only — no orientation, skew or mirror.

Production intake. ~33 min per 10k files on 4 cores. Trust the scanner operator's orientation; just clean and move on.

iqc-cli batch C:\scans C:\cleaned --mode daily
--mode thorough
~2-3 hrs
10k files · 4 cores

Thorough QA

Full pipeline + auto-mirror detection.

Periodic QA passes. Catches missed rotations and rare scanner-mirror failures across every step in the pipeline.

iqc-cli batch C:\scans C:\cleaned --mode thorough --from-flags suspect

The reviewer & editor

See every fix. Override any of them.

Open any page to compare the cleaned output against the original, side by side. Rotate, crop, whiteout or annotate by hand, flag it for another pass, or approve it — all keyboard-driven. The operator always has the final say.

A Mark page correct
2 Flag rotate-90
6 Flag suspect
W Whiteout tool
R Rotate 90° CW
⌃S Save & close
The InstaDocs reviewer and editor — cleaned output beside the original, with rotate, crop, whiteout and annotation tools. Document content blurred for privacy.

Live InstaDocs reviewer · documents & names blurred for privacy

Enterprise safety

Accountable to the archive it protects.

These documents are records of public consequence. Every automated decision stays traceable, reversible and explainable — even when that costs a little speed.

Auditable by design

Run history, per-file diagnostics and a manifest trace for every batch. You can always answer "what changed, when, and why."

Originals never lost

In-place QC writes a byte-identical backup first, sha256-verified. A bad batch is one restore command away — not a re-scan.

Offline & private

Per-machine, Ed25519-signed licensing verified offline. No phone-home, no cloud upload — runs on air-gapped machines.

Lean updates

Signed installer with delta updates — typically 5–15 MB versus a 133 MB full download — pulled from Cloudflare R2.

InstaDocs main window — file queue with per-page status, flags and a full run history panel. Document content and file names blurred for privacy.

Every run recorded — per-file status, flags and history. Documents & names blurred for privacy.

Licensing

Subscription or pay-as-you-go. Your call.

License InstaDocs as a subscription — billed monthly, annually, quarterly or half-yearly — or pay per page with no subscription commitment. Every plan is quoted to your volume and deployment.

Pay-as-you-go
Per page

No subscription commitment.

  • Full pipeline — daily + thorough
  • Operator GUI + CLI
  • Run history & restore
  • Pay only for pages processed
Get a quote

Pricing set per engagement · contact sales for a quote

Prefer to see it on your own files first? Send 50 pages — we'll run the full pipeline free, no commitment.

Questions

Before you download.

What file formats does it handle? +

Multi-page and single-page TIFF, and PDF — the formats document-archive scanning most commonly produces. Each cleaned file keeps its original format, alongside a per-file report.

Which languages does the OCR understand? +

25 in total — Tamil, English, Hindi, Bengali, Telugu, Kannada, Malayalam, Gujarati, Punjabi, Odia, Assamese, Urdu, Marathi, Nepali, Sinhala, Sanskrit, Arabic, Persian, Hebrew, Chinese (both scripts), Japanese, Korean, Thai and Vietnamese. English ships in the installer; the rest download on demand. InstaDocs detects each page's script automatically and applies the right language — no flag to set.

Does it run offline? +

Yes — entirely. There is no cloud upload and no phone-home. Licensing is Ed25519-signed and verified locally, so it runs on air-gapped machines.

Will it ever touch my originals? +

Only in opt-in in-place mode, and even then it writes a byte-identical, sha256-verified backup of every file first. Restore a single file or a whole run at any time.

How does licensing work? +

InstaDocs is licensed as a subscription — billed monthly, annually, quarterly or half-yearly — or pay-as-you-go per page, with no published price list since terms depend on volume and deployment. Contact us for a quote. Licenses are machine-bound, Ed25519-signed and verified entirely offline.

Can I roll it out across many machines? +

Yes. A silent/unattended installer supports fleet rollout, the CLI scripts batch jobs, and an optional Windows Explorer right-click menu drops QC into existing operator workflows.

What operating systems are supported? +

Windows is the supported operator target (signed installer, GUI + CLI). A Linux dev build exists for pipeline and automation use.

Get started

Put a clean, upright archive in front of every reader.

License InstaDocs as a subscription or pay-as-you-go per page — every plan is quoted to your volume. Download the installer and run a real batch today; nothing leaves your machine.

Compliance docs available on request.

PS> iex (irm https://get.appz.dev/instadocs/install.ps1)