Coming Soon

The models teaching machines to understand sign language.

Uvilox AI is built on three pieces: a recognition model for Indian Sign Language, a realtime agent that turns what it understands into speech, and an action layer that can actually place the call or send the message. All of it is in active development — this page explains how it fits together, honestly.

Model lineup

The models behind Uvilox AI.

None of these are publicly available yet. We publish scope and status only — no accuracy claims until we can share a benchmark we stand behind.

Coming Soon

UViLox Alpha

ISL fingerspelling · A–Z, 0–9

Frame-level landmark classifier for alphabets and digits. Used as the fallback path when a sign falls outside the word vocabulary.

v0.4 — internalNot public
Coming Soon

UViLox Lex

ISL words & short phrases

Sequence model over hand, pose and face landmarks. Trained on a small in-house Indian Sign Language set that we are still expanding.

v0.7 — internalNot public
Coming Soon

UViLox Voice

Sign → intent → speech

The layer that turns a recognised sign sequence into a spoken sentence and a real action — placing a call, sending a message, raising an SOS.

prototypeNot public
Public model

Try ISL letter recognition in your browser.

Our Indian Sign Language fingerspelling model is the first UViLox model available publicly. Show a letter to your camera or upload a photo, and the model tells you which letter it sees. Letters only for now — words and phrases are still in development.

Camera or photo
Live
Start the camera or upload a photo of an ISL letter.

Frames are sent to the UViLox recognition service only when you press a button, and are not stored by this page.

Model output
—
No letter detected yet

Supported vocabulary: ISL fingerspelling — letters A–Z and digits 0–9. Hold one clear hand shape in frame, with good lighting and a plain background, for the most reliable reading.

Realtime agent

Give an instruction. Watch the agent act on it.

A guided walkthrough of how the UViLox realtime agent is designed to work on LiveKit: audio and sign frames stream into a room, the model reads intent, and a tool call places the call, sends the message, or raises the alert. This walkthrough is scripted — the live agent is still in development.

Try an instruction
Coming Soon
livekit · room: uvilox-demo
idle
  1. LiveKit room
    Joining realtime audio room
  2. Capture
    Streaming mic / ISL frames
  3. Understand
    Transcribing and reading intent
  4. Tool call
    Executing the requested action
  5. Speak
    AI voice replies out loud
Transcript

Waiting for input…

Tool call

no tool invoked yet

Simulated walkthrough for illustration. No call, message, or alert is sent from this page, and no microphone is accessed.

Architecture

How a sign becomes a call, a message, or an alert.

A plain-language walk through the stack we're building. Numbers and benchmarks will follow once the models are stable enough to publish them.

Stage 01
Landmark pipeline

Hand, pose and face keypoints

Instead of raw video, we work on normalised keypoints. It keeps inference light enough for mid-range Android phones and means we never need to ship someone's face to a server to understand a sign.

  • Keypoints extracted on-device
  • Normalised for distance and camera angle
  • Works in variable indoor lighting
Stage 02
Temporal model

Sequence decoding for continuous signing

Signs are motion, not stills. A temporal model reads a rolling window of keypoints and emits a phrase, with the fingerspelling model handling anything outside the word vocabulary.

  • Rolling-window sequence decoding
  • Fingerspelling fallback for unknown words
  • Confidence gating before anything is spoken
Stage 03
Realtime transport

LiveKit rooms for voice and captions

Recognised phrases become an agent turn inside a LiveKit room — the same room that carries the call audio — so the AI voice, live captions, and the person on the other end stay in sync.

  • One room for audio, captions and agent turns
  • Tool calls for calling, messaging and SOS
  • Barge-in and turn handling handled by the agent
Stage 04
Safety layer

Nothing acts without a confirmation path

An assistant that can dial numbers needs guardrails. Actions are previewed before they run, emergency flows are separated from everyday ones, and the user keeps control of contacts and permissions.

  • Action preview before execution
  • Separate emergency escalation path
  • User-controlled contacts and permissions