# LAVA

> AI sign language anonymization

- **Slug:** `lava`
- **Organization:** SIL Global
- **Website:** https://lava.multilingualai.com
- **HTML page:** https://www.tools.bible/tools/lava
- **Markdown page:** https://www.tools.bible/tools/lava.md

## Classification

- **Translation stages:** Bible Publishing, Scripture Engagement
- **Medium:** Sign Language, Video
- **Pricing:** Free (Restricted)
- **Release status:** General Availability
- **Localization:** English
- **Translation type:** Any
- **License:** Creative Commons (various)

## Description

LAVA (Language-Agnostic Video Anonymization) is a research and development project that uses AI to anonymize sign language videos by changing the appearance of the signer while preserving the signing, timing, and translated message. Since sign language Scripture is published as video, the identity and appearance of the signer are part of the translation itself. This can create security risks in sensitive contexts and can also affect Scripture engagement when viewers personally know the signer or when older videos feel dated because of clothing, hairstyles, or other visual elements.

LAVA is designed to help teams use existing sign language Bible translation videos without having to re-record the translation. A team can submit an ordinary video and receive an anonymized version in which the signer looks different, while the original signed content remains intact. This is especially valuable for protecting translators whose identity should not be visible in publicly distributed Scripture videos.

A goal of the project is to support full-video anonymization, not just isolated signer footage. Many sign language publications include complex backgrounds, illustrations, slides, captions, or on-screen text. LAVA aims to preserve those elements while changing only the signer’s appearance, so teams do not have to separate the signer from the rest of the publication, anonymize the signer separately, and then manually reassemble the final video.

LAVA takes a software-based approach and is designed to work with ordinary video, including material recorded years ago. This complements projects like Chameleon, which use specialized multi-camera recording and 3D avatar workflows. LAVA’s advantage is that it can fit into existing recording and publishing workflows and be applied retrospectively to previously produced Scripture videos. As with Chameleon and other existing anonymization tools, LAVA cannot guarantee 100% anonymization, for example, in videos where a uniquely identifying signing style is used.

LAVA is currently being developed and tested in partnership with sign language translation teams and organizations including 55 Table, SLTT, DOOR International, and Deaf Harbor. The long-term goal is to make LAVA available as a shared service through an API and integrations with existing sign language translation tools, giving teams access to advanced video anonymization without requiring their own AI specialists or expensive computing infrastructure.

Learn more about SIL AI projects at [ai.sil.org](https://ai.sil.org/).

## Screenshots

![LAVA screenshot](https://www.tools.bible/api/media/file/screenshot1.png)

![LAVA screenshot](https://www.tools.bible/api/media/file/screenshot2.png)

---

Source: [Tools.Bible](https://www.tools.bible) · Catalog: [https://www.tools.bible/tools.md](https://www.tools.bible/tools.md) · Index: [https://www.tools.bible/llms.txt](https://www.tools.bible/llms.txt)