Vahini StudioHandwriting Analysis Demo
1Details 2Mode 3Capture 4Report
Step 1 · Who is this for?

Tell us about the writer

Choose a role and add a few details. The finished report adapts its tone, imagery and sign-off to match.

Student
Kid-friendly & motivating
Parent
Warm progress view
Coach
Notes & sign-off
School
Institutional report
Individual
Self-paced adult
Click to
upload
PNG or JPG. Shown top-right of the report cover.
Nothing leaves your browser - the demo runs entirely on-device.
Step 2 · How was it written?

Choose a capture mode

Vahini works two ways. Analyse an existing photo or scan, or capture live with the patented IMU sensor pen for the full picture - including pressure and pen dynamics a photo can’t see.

Photo / Scan

Upload a picture of handwriting. The deterministic-CV engine binarizes, segments and measures the page.

  • Works with any phone photo or scan
  • ~14 factors measured directly
  • Dynamics estimated by proxy
Choose Photo / Scan
Patent 584433

Vahini Writing Assistant

Write live with the IMU sensor pen. 19 axes stream at 208 Hz - force, tilt, velocity and pen-lifts in real time.

  • Dual LSM6DSO + magnetometer + force sensor
  • Kalman-filtered live signals
  • Dynamics factors measured, not proxied
Launch the pen
No dataset is collected - the pen stream is a live simulation for this demo.
Step 2 · The handwriting sample

Upload & choose the passage

Add a photo or scan of the handwriting, and tell us which passage was copied. The engine measures the actual marks against that model.

Drop a handwriting photo

or click to browse. A clear, well-lit shot of a few lines on plain or ruled paper works best.

Reference sentence (optional with recognition)

With the local PP-OCRv5 recognition server running (python ppocr-server.py), the written text is recognised automatically and this reference isn’t needed. Offline, we compare against the sentence below.

Step 3 · Vahini Writing Assistant

Writing - sensors live

Hold the pen and write naturally. The 19-axis array streams below. When you’re done, press Finish to generate the report.

REC · 208 Hz
Rear IMU Magneto Front IMU Force
Live ink · reconstructed path
Tip force-
Pen tilt / slant-
Velocity-
Samples0
Strokes0
Pen-lifts0
KalmanON
Step 3 · The pipeline

Analysing the handwriting

Binarize → segment → text detection & recognition → deterministic-CV measurement → aggregate. Exactly the architecture from the engine spec.

Input

Pipeline

Vahini 20-Factor Engine · v3.0
Recognition runs on the Vahini server when configured via window.VAHINI_OCR_ENDPOINT; otherwise the in-browser detector keeps the demo fully offline. All geometry is computed on-device.