Educational Sign Language Translator
Plan tokens from text or speech and render SiGML with CWASA
Text
Language hint (for STT only)
auto
en
ru
es
de
fr
zh
ja
Plan (spaCy)
Plan (OpenAI)
Plan
STEM / EDU tools
Use EDU NLP pipeline
Apply abbreviations
STEM mode
Load abbreviations JSON (URL)
Load
Or upload JSON file
Clear
EDU NLP extracts pronouns, verb lemmas, and key nouns; abbreviations collapse multi-word forms (e.g., “carbon dioxide” → “co2”).
Microphone
● Start recording
Transcribe
Idle
On-device Whisper (transformers.js). First run downloads model; allow ~100–200MB cache.
Video Upload → Sign
Video → Sign
No file
Settings
OpenAI API key (optional, stored locally)
Save
Database .sigml URL(s)
Load
You can load multiple files one-by-one. Loaded signs override earlier duplicates by gloss.
Concept CSV (optional)
Load
Or upload CSV/TSV
Planner
Planner log
Normalized
Tokens
Vocabulary (from database)
EDU Diagnostics
Timeframe guess
Abbreviations applied
Render (CWASA)
Play planned signs
Stop
CWASA loading…
Queued tokens
Renders by concatenating the matching
<hns_sign>
blocks into one SiGML and sending to
CWASA.playSiGMLText
.