Complete system
From camera frame
to spoken page.
Seven detailed components cooperate across perception, recognition, decision, and interaction. Each page documents inputs, methods, outputs, failure cases, and current limits.
Camera capture
Turns a live camera stream into stable frame candidates the rest of the pipeline can evaluate.
Open technical page ↗02page boundsPage and text detection
Separates the printed page and its main text region from the camera background.
Open technical page ↗03clean cropImage enhancement
Improves the isolated text crop while monitoring blur and glare that filters cannot recover.
Open technical page ↗04median frameMedian burst stacking
Combines several nearby frames to reduce temporary variation before recognition.
Open technical page ↗05decoded textOCR recognition and validation
Generates candidate text, checks its plausibility, and avoids accepting recognition blindly.
Open technical page ↗06audio outputSpeech output
Cleans accepted text and turns it into paced, interruptible spoken output.
Open technical page ↗07feedback loopSpoken user guidance
Turns measurable image problems into short instructions that help improve the next frame.
Open technical page ↗