Purpose
Acquire readable frames without assuming the user can visually judge the preview.
Role in the complete pipeline
Capture is the first quality gate. Blur, distance, exposure, and framing problems introduced here propagate into every later stage.
INPUT
→A live camera feed and camera metadata.
METHOD
- Acquire frames continuously
- Measure sharpness and exposure
- Retain a short burst of candidates
- Prefer stable frames for downstream processing
A frame or short burst, plus quality signals used by detection and guidance.
Common failure cases
- Handheld motion merges character strokes
- Text is too small at the current distance
- Glare removes visible detail
- The page falls outside the frame
How VisoraAI responds
- Ask the user to hold steady or move closer
- Delay capture until stability improves
- Pass several frames to burst stacking
Current limitations
- Quality thresholds vary by camera and lighting
- Automatic capture can feel slow when conditions keep changing
Planned improvements
- Calibrate thresholds per camera
- Add exposure-aware capture
- Test automatic capture timing with users