Project Akshara

Everyone photographed the page.
Nobody recorded the hand.

Recording the hand · Telugu వాహిని MOTION · 208 SNAPSHOTS/SECOND
pressure · pace · hesitation · flow

The dataset nobody has built

World-first for Indic scripts

The pressure, pace and hesitation behind every stroke has never been recorded at scale, and for Indic scripts, barely at all. Whoever owns that motion owns the model. We decided it should be us, starting in India.

World firsthandwriting motion, recorded at scale, for Indic scripts
Five scriptsTelugu, Kannada, Malayalam, Tamil & English, live in the demo above
With schoolscollected where handwriting lives, on ordinary paper
Years, by designreal hands can't be rushed, scraped or bought
WHAT A SCAN KEEPS అ one frozen shape WHAT THE PEN KEEPS అ shape + how the hand made it

Why it doesn't exist yet

A photo keeps the shape.
Motion keeps the story.

A photograph shows what a letter ended up looking like. The recording shows the hand slow before a tricky loop, press harder on a retry, hesitate at the same joint twice. That second layer, real hands, real paper, in the scripts research forgot, is what Akshara collects.

the corpus, page after page whoever starts later, starts here school 1school 12school 40

A marathon, not a sprint

Years are the moat money can't buy.

Every sample is written by a real hand, page after page, school after school. It can't be scraped, rushed or bought. Granted hardware IP (Patent No. 584433), a head start measured in years, and school trust earned consent-first, three layers, each protecting the others.

Why it compounds →
a normal page, written in class consent gate opt-in · withdrawable identity separated the corpus quality-checked · balanced coaching reports back to students

How collection works

Simple on the surface.
Careful underneath.

  • Schools write, as they already do. Ordinary paper, normal class time.
  • Consent first, always. Opt-in, guardian sign-off, withdrawable any time.
  • Motion recorded, identity separated. Pressure and pace join the corpus; names don't.
  • Contributors benefit immediately. Coaching reports from day one, quality-checked as it grows.

What the corpus makes possible

One dataset. Four industries.

Handwriting that types itself

Write in your script, get digital text back. Doesn't exist for Indic scripts yet, Akshara is how it gets built.

Learning, seen early

Struggles with writing show up in motion long before they show up in grades.

Assistive writing

Real help for hands that shake or tire: writing that stays legible, progress you can measure.

A research foundation

Curated releases so labs can finally study writing the way it happens: in motion, on paper.

Responsible by design

Built the slow way, because it's the right way

Children are involved, so we set the bar ourselves before anyone asked: plain-language consent, guardian sign-off for minors, motion separated from identity, and the right to withdraw at any time. A dataset built with schools only works if schools can trust it for years.

Be part of the marathon

Run a school? Research handwriting or motor skills? Invest in companies that build patient, compounding assets? We should talk.

Reach out Investor overview