SignchatLooking for Signchat for Teams?Learn more
Signchat
    • Signchat Web
    • Signchat Bridge
    • ASL Classifier
    • Sign Pipeline
    • Architecture
  • Enterprise
  • Education
  • Pricing
Start a call

Sign with your hands,
they hear your voice

Real-time ASL-to-voice and live captions. Free, in your browser, no install.

Start a call
Custom ASL classifier · runs in browser
Top-k tokens from the last window
PIZZA · 0.91ME · 0.78LIKE · 0.84
Two seconds of silence → stitch
“I like pizza.”
Approve Edit
type instead — chat composer fallback
MediaPipeONNX RuntimeGeminiElevenLabsLiveKit
Sign-end to first audible byte · less than 1 s at P50
Review before broadcastauto · 2.0 s silence
“I like pizza.”

Stitched from your tokens by Gemini. The other side hasn’t heard anything yet — you have the final say.

PIZZAMELIKE
ApproveEdit↵ to send
Browser-direct · no relay
Live captionshearing peer · streaming
Hearing peerOkay, what kind?
YouI like pizza.
Hearing peerPepperoni or margherita?
YouPepperoni.
Hearing peerComing up.
Bridge → FaceTime · Zoom · Meet · Teams

Signchat Bridge connects to the tools your team already uses

  • Zoom
  • Microsoft Teams
  • Google Meet
  • Slack
  • Discord
  • Webex
Why we exist

Interpreter math doesn’t work

About 500,000 Americans speak ASL as their primary language. The US has roughly 10,000 certified ASL interpreters, while most of them are scheduled into healthcare, legal, and government calls. For casual video calls with a friend, family member, or coworkers, the deaf side has been left out.

  • ~500KAmericans whose primary language is ASLNIDCD
  • ~10Kcertified ASL interpreters in the U.S.Language Services Associates
  • <1 secsign to spoken voice — less than 1 second at P50

Signchat fills the gap.

Start a call

Sign to Chat
Sign to voice

Sign with your hands. They hear your voice.

Your camera feeds a custom ASL classifier model running locally in your browser with sub-20 millisecond latency. Recognized signs are stitched into fluent sentences by a fine-tuned LLM, streamed back as natural synthetic speech, and mixed with your microphone into one stable voice channel. To the receiving person, it sounds like you are talking.

How sign-to-voice works
  • Custom ASL classifier runs locally — your camera feed never leaves your machine
  • Sign-end to first audible byte: less than 1 second at P50
  • Synthetic speech mixed with your real microphone on one stable voice channel
  • No interpreter, no install, no account
PIZZA·ME·LIKE
“I like pizza.”
Signchat

Sign-language ↔ voice video chat that runs in your browser. No install, no backend, ~1-second end-to-end.

BeaverHacks 2026 · v0.1.0

Product

  • Start a call
  • Architecture

Open source

  • GitHub repo
SIGNCHAT