Purpose
Deliver recognized information in a form the user can follow without reading a screen.
Role in the complete pipeline
Speech is the primary output channel. It also coordinates with guidance so corrective prompts do not collide with document reading.
INPUT
→Validated OCR text and reading controls.
METHOD
- Normalize whitespace and punctuation
- Split long passages into manageable units
- Queue speech without overlap
- Pause for higher-priority guidance
Spoken document audio with a controllable reading sequence.
Common failure cases
- OCR artifacts sound confusing
- Long output becomes hard to follow
- Guidance interrupts at the wrong time
- Punctuation produces unnatural pacing
How VisoraAI responds
- Suppress rejected OCR
- Chunk the passage
- Separate guidance and reading queues
Current limitations
- Text structure is only as good as OCR output
- Voice controls are still a planned interaction
Planned improvements
- Add repeat, pause, and slower-reading commands
- Detect headings and lists
- Test pacing preferences