Purpose
Close the loop when the user cannot visually verify camera position or image quality.
Role in the complete pipeline
Guidance interprets blur, glare, distance, page visibility, and stability signals before recognition is attempted again.
INPUT
→Quality measurements, detection state, OCR validation, and prompt history.
METHOD
- Prioritize the most actionable issue
- Map it to a short spoken instruction
- Apply cooldowns to prevent repetition
- Confirm when the page is ready
Prompts such as “Move closer,” “Hold steady,” “Reduce glare,” or “Ready to read.”
Common failure cases
- Prompts repeat too often
- Several problems compete for attention
- A wrong diagnosis makes capture worse
- Instructions arrive too late
How VisoraAI responds
- Speak one issue at a time
- Use prompt cooldowns
- Stop prompting once the frame is acceptable
Current limitations
- Thresholds need user testing
- Not every failure can be translated into a simple correction
Planned improvements
- Evaluate prompt clarity with users
- Tune priorities by failure type
- Add nonverbal alignment cues where appropriate