Skip to content

AI-powered real-time sign language accessibility proposal for media, education, and public communication. Public-good licensed

License

Notifications You must be signed in to change notification settings

dodo477/DeafVisionAI

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

11 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

DeafVisionAI – A Public Proposal for AI-Driven Sign Language Accessibility

Status: Concept Proposal – Open to Collaboration
License: GNU Affero General Public License v3.0 (AGPL-3.0)
Proposed by: Dominik Dostál
Contact (optional): https://github.com/dodo477


🔍 Overview

DeafVisionAI is an open-source proposal to create a real-time or pre-rendered AI accessibility layer that transforms spoken digital content — such as movies, podcasts, lectures, and livestreams — into native sign language through expressive, culturally aware avatars.

Unlike subtitles, which force Deaf users to rely on written translations of spoken grammar, this system aims to deliver fluent, immersive, and emotionally accurate sign language representations, matching both linguistic structure and tone. It can be deployed as a browser plug-in, API layer, or platform-side overlay — and is designed to work across YouTube, Netflix, Spotify, Zoom, and more.


🎯 Vision

This is not a commercial product.
It is a public-good blueprint — open for developers, researchers, NGOs, and accessibility advocates to build upon freely.

The goal is simple:

Make digital spoken content fully accessible in sign language, anywhere, anytime.


⚠️ Why Subtitles Are Not Enough

  • Many Deaf users consider sign language their first language — not written text.
  • Subtitles often miss emotional tone, sarcasm, urgency, or cultural nuance.
  • Reading subtitles requires visual split-attention, reducing immersion and accessibility.
  • For audio-only media (e.g., podcasts), subtitles are not even an option without separate transcripts.

🧬 What This Project Synthesizes

Several promising efforts exist, but none offer all of the following in one integrated, open system:

Feature Existing Projects DeafVisionAI
Browser plug-in for real-time sign overlays ✅ SignSync, SignUp ✅ Planned
Multi-platform support (YouTube, podcasts, Netflix) ❌ Partial ✅ Core goal
Native sign language grammar handling (not word-for-word English) ⚠️ Basic ✅ Prioritized
Avatar customization (ethnicity, style, dialect) ❌ Mostly fixed ✅ Included
Open-source licensing with anti-enclosure protection ⚠️ Some Apache, some closed ✅ AGPL-3.0
Designed as a public good with no corporate lock-in ❌ Rare ✅ Explicit mission

🔧 Key Components

  • Speech-to-Text Engine (e.g. Whisper)
  • Sign Language Grammar Translator (ASL, BSL, etc.)
  • AI Gesture & Facial Expression Synthesizer
  • User-side Overlay Plug-in / API Integration
  • Customizable Signer Avatars (race, signing style, expressiveness)

📦 Potential Use Cases

  • Watching Netflix or YouTube with AI-signed dialogue
  • Experiencing podcasts or interviews in native sign language
  • Following livestreamed lectures, webinars, or Zoom calls
  • Receiving public broadcasts and emergency messages accessibly
  • Supporting Deaf/hearing hybrid households with synchronized audio + sign

See SYSTEM_SKETCH.md for a full technical and strategic breakdown.


🤝 Call for Collaboration

This project is being published as a proposal, not a finished product.
I do not intend to build this myself due to other high-priority commitments — but I believe it should exist, and I invite others to take this idea forward.

You are free (and encouraged) to:

  • Fork this repo and begin prototyping
  • Use it to apply for grants, EU funding, or institutional support
  • Adapt the vision to local sign languages or platforms
  • Expand it into formal specs or research proposals

No permission needed. Attribution appreciated.


🔐 License

This project is published under the GNU Affero General Public License v3.0 (AGPL-3.0).
This ensures:

  • It remains a public good.
  • Any hosted service (e.g. SaaS, plug-in) based on this must also share its source.
  • Commercial services can exist — but must remain open, share-alike, and accessible.

See LICENSE for full terms.


🧭 Ethical Use Statement (Non-Binding)

DeafVisionAI is proposed as a public good intended to expand access, dignity, and linguistic autonomy for Deaf communities through open, non-exploitative technology.

While this project is licensed under the AGPL-3.0, which ensures legal openness, we also request that developers and institutions using or adapting this project respect the following ethical principles:


⚖️ Ethical Principles

  1. Inclusion, Not Replacement

    • This system is not intended to replace human sign language interpreters.
    • It should be used where no interpretation is available, or to supplement access—not to displace cultural labor.
  2. Representation Matters

    • Deaf individuals, sign language users, and culturally Deaf advisors should be consulted in development and deployment.
    • Regional dialects, racial identity, and signing styles must be respected and represented.
  3. No Closed Commercialization

    • Do not build closed-source or proprietary systems from this work without open, shared improvements.
    • Monetization may occur only in ways that preserve public access and transparency.
  4. No Surveillance or Coercion

    • This project may not be used for surveillance, behavioral profiling, or extracting biometric data from Deaf users.
    • It must not be deployed in authoritarian or coercive contexts without consent and oversight.
  5. Cultural Sensitivity

    • Sign languages are not universal. Avoid treating them as interchangeable.
    • Emotional tone, rhythm, and performance style are integral to communication and must not be flattened or sanitized.

🤝 This Is a Civic Request, Not a Legal Clause

I understand that open projects invite adaptation.
This is not a restrictive license or formal barrier — it is a conscience clause for ethical developers and institutions.
If you benefit from this project, consider aligning with its intent:

To empower Deaf people, not merely “serve” them.


🧠 Related Projects and Inspirations

These existing efforts helped inform the shape of DeafVisionAI:

DeafVisionAI aims to synthesize and generalize these ideas into a flexible, scalable, and culturally respectful open system — built by and for the community.


✍️ How to Get Involved

If you’re:

  • A developer interested in building the prototype
  • A researcher working in sign linguistics or accessibility
  • A Deaf community member who wants to shape this
  • A funder or nonprofit looking for impactful projects

Then this blueprint is for you.

Open an issue, fork the repo, or email me. Let’s make this real.

About

AI-powered real-time sign language accessibility proposal for media, education, and public communication. Public-good licensed

Topics

Resources

License

Contributing

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published