What is document autocapture?

The autocapture mechanism in our SDKs ensures that documents are captured at the highest possible quality. During a capture session, the SDKs continuously attempt autocapture. However, for the first 10 seconds of a document capture attempt, the user cannot manually trigger a capture. If the SDKs can’t complete autocapture within this 10-second window, a manual-capture fallback appears, providing a button for users to capture the document themselves.

Document autocapture flag

We strongly advise leaving autocapture enabled except in the specific scenario described below.

We have observed an isolated issue in which certain external dependencies in a partner’s app cause crashes when transitioning to manual capture. This occurred on a partner’s app using the v10 Flutter SDK with separate capture components. The exact cause remains unknown.

Option
Behavior
Manual Capture
Auto Capture
Passrate Impact

autoCapture

Auto-capture with manual fallback (10s)

✅ fallback

Optimal

autoCaptureOnly

Only auto-capture

⚠️ Possible drop

manualCaptureOnly

Only manual capture

⚠️ Possible drop

Note: Only set autoCapture: autoCaptureOnly if you encounter this specific crash, and be aware of the performance differences between autocapture and manual capture.

Manual Capture Fallback time

If you wish to extend or reduce time based on your internal test or requirement, you can use the autoCaptureTimeout flag.

Note: We generally advise not changing this unless you absolutely need to.

Performance of auto-capture vs. manual capture for documents

Based on millions of usage data and tests, auto-captured documents have significantly higher pass rates than manually captured ones—especially on mid-range devices. Two key factors contribute to this:

  1. Optimized processing pipeline Our autocapture pipeline is designed to minimize blur and maximize document quality. A document is captured only after all required quality checks have passed.

  2. Faster shutter response Autocapture uses a live video stream to evaluate quality. Once the system decides to capture, the delay from decision to shutter is under 100 ms—much faster than a human pressing a button. Even a small time lag or the slight hand movement involved in manual capture can introduce blur.

Limitations of auto-capture

Autocapture may struggle in certain environments; manual capture is retained as a fallback by design. Common challenges include:

  1. Multiple ID-shaped objects Object-detection models first locate the ID within the frame. If there are several ID-shaped items (e.g., multiple cards, open wallets, small booklets) in the background, the model can have difficulty focusing. A plain background improves the pass rate.

  2. Poor lighting Low-light conditions challenge all capture methods, reducing image clarity and model confidence. This has a negative impact on success rate for both autocapture and manual capture.

  3. Limited device stabilization Excessive movement of the user’s device degrades image quality. This affects both autocapture and manual capture, as motion blur increases with movement.

Last updated

Was this helpful?