OCR Capture errors

Root Causes – In Order of Likelihood

1. Screen / Window Focus Lost Before OCR Capture (Most Common)

The OCR Capture (Area) action requires the target window/screen to be in focus and fully rendered at the moment of capture. Inside a larger bot:

• A previous action (click, keystroke, database call, file operation) may have shifted focus away from the target window.
• The window may be partially obscured, minimized, or behind another window.
• A previous action may have triggered a loading/transition state where the page is not fully rendered when OCR fires.

Fix:

• Add a Window: Focus or Window: Restore action immediately before the OCR Capture step.
• Add a Wait/Delay (2–5 seconds) before the OCR action to ensure the page has fully loaded/updated.
• Use Window: Exists to confirm the target window is active before OCR fires.

2. Race Condition – OCR Fires Before Page Has Fully Updated

Since you are using OCR specifically to verify a page has updated properly, the OCR may be firing before the update is complete:

• The previous action (submit, save, navigate) triggers a page update.
• The bot immediately proceeds to the OCR step.
• The page is still rendering/loading when OCR tries to initialize.
• The OCR engine fails to capture a stable screen state → "Unspecified error".

Fix:

// Before OCR Capture:
Action: Wait (3–5 seconds)
  OR
Action: Wait for Window (wait for the updated page title/element to appear)
  OR
Action: Image Recognition: Wait for image to appear (wait for a known 
        element on the updated page)
  THEN
Action: OCR Capture (Area)

3. OCR Engine Resource Exhaustion / Memory Leak in Long-Running Bot

The OCR engine (Tesseract or ABBYY, depending on your AA version) initializes a separate process/engine instance each time it is called. In a larger bot:

• If the bot has been running for a long time before reaching the OCR step.
• If there are many previous actions consuming memory (large DataTables, many file operations, multiple browser sessions).
• The OCR engine may fail to allocate sufficient memory to initialize, resulting in "Unspecified error during initialization."

Fix:

• Add an explicit Wait before the OCR step to allow system resources to stabilize.
• Check Task Manager during bot execution to monitor memory usage at the point of OCR failure.
• If memory is high, consider breaking the bot into smaller child bots called via Run Bot to release memory between phases.

4. Screen Lock / Session State Change

In a larger bot, especially one that runs for extended periods:

• The Windows session may have locked or gone into a screen saver state before the OCR step.
• A UAC prompt or system notification may have appeared on top of the target window.
• A Windows Update notification or background process may have overlaid the screen.

OCR Capture (Area) captures the literal screen pixels — any overlay or lock screen will cause it to fail or capture the wrong content.

Fix:

• Ensure the Bot Runner session is configured to prevent screen lock during automation.
• Add Window: Focus on the target application immediately before OCR.
• Check for and dismiss any system overlays before OCR fires.

5. Previous Action Left a Modal Dialog / Popup Open

A previous step in the larger bot may have triggered a modal dialog, popup, or alert that:

• Is not visible (hidden behind the main window).
• Is blocking the OCR engine from accessing the screen properly.
• Causes the OCR initialization to fail with an unspecified error.

Fix:

• Add error handling before the OCR step to check for and dismiss unexpected dialogs.
• Use Window: Close or Keyboard: Press Escape to dismiss any lingering popups before OCR fires.

6. OCR Package / Tesseract Engine Temp File Conflict

The AA OCR engine writes temporary files during initialization (language data files, engine config). In a larger bot:

• If a previous OCR call in the same bot (or a previous bot run that didn't clean up) left locked temp files.
• The OCR engine fails to initialize because it cannot write its required temp files.
• This only manifests in the larger bot if there was a prior OCR call earlier in the same execution.

Check:

• Look in C:\Users[BotRunner]\AppData\Local\Temp\ for any AA OCR-related temp files.
• Check if there is another OCR action earlier in the same bot that may not be releasing the engine properly.

Fix:

• Add a Wait between consecutive OCR calls in the same bot.
• Ensure OCR actions are not running in parallel threads within the same bot.

7. Capture Area Coordinates Are Off-Screen or Invalid in Context

The OCR Capture (Area) uses fixed screen coordinates recorded at design time. In the larger bot:

• A previous action may have changed the screen resolution or window size/position.
• The target application window may have opened at a different position than when the bot was designed.
• The capture area coordinates may now point to an off-screen or invalid region.

Fix:

• Use relative coordinates or anchor-based capture rather than absolute screen coordinates.
• Add a Window: Move/Resize action before OCR to ensure the target window is in the expected position.
• Verify the capture area is valid by adding a Screenshot action just before OCR to visually confirm what the screen looks like at that point.

Recommended Action Plan

Debugging Technique: Isolate the Context

To pinpoint exactly what is causing the conflict, use this approach:

// Binary search the larger bot:

Step 1: Run the first HALF of the bot → then run OCR → does it fail?
         YES → problem is in first half
         NO  → problem is in second half

Step 2: Narrow down further until you identify the SPECIFIC action
        that, when run before OCR, causes the failure.

Step 3: That action is your root cause.

This will definitively identify which preceding action is corrupting the state that OCR depends on.

Pro Tip: Alternative to OCR for Page Verification

Since you are using OCR specifically to verify a page has updated, consider these more reliable alternatives:

These approaches are more reliable than OCR for page verification as they don't depend on screen rendering state.

References

• AA OCR Package Docs: https://docs.automationanywhere.com/bundle/enterprise-v2019/page/enterprise-cloud/topics/aae-client/bot-creator/commands/ocr-package.html
• AA Image Recognition: https://docs.automationanywhere.com/bundle/enterprise-v2019/page/enterprise-cloud/topics/aae-client/bot-creator/commands/image-recognition-package.html