vidaio-subnet

<div align="center">

Vidaio Subnet: Revolutionizing Video Processing with AI-Driven Decentralization <!-- omit in toc -->

Please check our X account to follow us. Website

---

</div>

Table of Contents

1. Introduction
2. Subnet Architecture
   • Overview
   • Miners
   • Validators
   • Synapses
     • Synthetic Query
     • Organic Query
3. Setup
4. Roadmap
5. Appendix
   • Technical Glossary
   • References
   • Contact Information

1. Introduction

Vidaio's mission is to democratise video enhancement through decentralisation, artificial intelligence, and blockchain technology. Leveraging the Bittensor ecosystem, Vidaio provides creators, businesses, and developers with scalable, affordable, and high-quality video processing solutions including upscaling and compression, while ensuring full ownership and control over their content.

2. Subnet Architecture

2.1 Overview

• Miners: Handle video processing tasks including upscaling and compression, optimizing models to ensure high-quality outputs.
• Validators: Assess miner performance using predefined metrics to maintain network integrity across all video processing workflows.

2.2 Miners

Miners enhance video quality and optimize file sizes using AI-driven processing techniques. They can:

• Optimise open-source models or develop proprietary ones for superior upscaling and compression results.
• Handle video upscaling and compression requests from validators and end-users.
• Process both upscaling tasks (enhancing video quality) and compression tasks (reducing file size while maintaining quality).

2.3 Validators

Validators ensure miners deliver consistent, high-quality results by evaluating performance through synthetic and organic queries for both upscaling and compression workflows.

2.4 Synapses

2.4.1 Synthetic Query

Validators benchmark miner performance using controlled datasets:

• Upscaling: Downscale a high-resolution video to low-resolution, then miners upscale it back to high resolution.

• Compression: Provide high-quality videos for miners to compress while maintaining optimal quality-to-size ratios.

• Validators assess the processed outputs using metrics VMAF and PieAPP for quality evaluation.

2.4.2 Organic Query

Real-world video data uploaded by users is processed as follows:

• Videos are chunked and queued for miners.
• Miners process and apply upscaling or compression based on user requirements.
• Results are aggregated and delivered back to users.

2.5 Incentive mechanism

• Incetive Mechanism Guide

3. Setup

• Validator Setup Guide
• Miner Setup Guide
• Inference Miner Checklist
• Competition Miner Checklist

4. Roadmap

Phase 1: Implementing the Video Processing Synapses

• Launch the subnet with AI-powered video upscaling and compression capabilities.
• Focus on real-time processing of videos for both quality enhancement and size optimization.

Phase 2: Developing Advanced Video Processing Models

• Build AI models for adaptive bitrate streaming and intelligent compression.
• Optimize bandwidth usage while maintaining video quality across different use cases.

Phase 3: Implementing the Transcode Optimization Synapse

• Introduce AI-driven transcoding for compatibility across devices.
• Evaluate miners on speed, quality, and efficiency for all processing workflows.

Phase 4: On-Demand Streaming Architecture

• Enable decentralized on-demand video streaming with integrated storage.
• Utilize peer-to-peer (P2P) models for redundancy and high availability.

Phase 5: Live Streaming Through the Subnet

• Introduce live streaming with real-time AI-powered upscaling, compression, and transcoding.
• Integrate adaptive bitrate streaming for smooth playback.

Phase 6: Subnet API for Real-World Integration

• Develop a RESTful API for seamless integration with external platforms.
• Include features for uploading, processing, and retrieving videos with multiple processing options.

5. Appendix

A. Technical Glossary

• VMAF: Video Multimethod Assessment Fusion
• Video2x: Vidio upscaling model
• TOPIQ: Top-down Image Quality Assessment
• LPIPS: Learned Perceptual Image Patch Similarity
• Bittensor Subnet: Decentralized AI Framework