A framework for decentralized AI agent translation and storage encryption using GaiaNet and Nillion Encrypted Storage
NativeEcho is a decentralized translation application that allows users to translate YouTube videos into their native language using AI-powered language models. This project integrates Python for web scraping YouTube closed captions, JavaScript for data uploading, and a decentralized storage layer on Nillion. The AI translation is powered by a language model deployed on a GainaNet node running on AWS EC2.
NativeEcho is a decentralized translation application designed to bridge language barriers for educational content by leveraging LLMs on GainaNet Nodes and decentralized storage. At its core, we built a backend that scrapes YouTube captions using YouTube Data API v3 and other packages and stored them in Nillion's Encrypted vault according to a custom schema.json we defined specially for the use case. The extracted data is then securely stored in Nillion Encrypted Storage, allowing decentralized access to video metadata. Something worth noting is that this project is in response to the ETHGlobal's desire to translate their Youtube videos into many languages, marked by them as a HIGHLY important bounty (https://ethglobal.notion.site/Earning-ETH-Bounties-for-Staking-acf5a3ff10b4436982c234d376dcb096)