Subnet 17
Three Gen
404
Three Gen enables users to create 3D worlds and games through simple descriptions and AI models
SN17 : Three Gen
| Subnet | Description | Category | Company |
|---|---|---|---|
| SN17 : Three Gen | Text-to-3D model | Generative Al | 404 |
Bittensor Subnet 17 is empowering individuals to create worlds and games through simple descriptions, thereby democratizing 3D content creation, enabling anyone to develop virtual worlds, games, and AR/VR/XR experiences. This subnet utilizes a diverse array of existing Open Source 3D generative models, including Gaussian Splatting, Neural Radiance Fields, 3D Diffusion Models, and Point-Cloud approaches.
By integrating these models into decentralized incentive-based networks via Bittensor, the subnet aims to foster innovation. Its goal is to initiate a gaming revolution centered around AI-native games, utilizing the broader Bittensor ecosystem to enable real-time generation of assets, voice, and sound. This approach empowers creative individuals, even those without coding or game development experience, to simply describe their game ideas and see them materialize instantly.
404 provides a software application known as the Service, allowing Users to mint and/or acquire unique, non-fungible tokens (NFTs) on third-party blockchains (the “Blockchain Platform”) using smart contracts. These NFTs, collectively referred to as Tokens, are associated with products, services, and benefits. To participate in Token purchases, Users must link a Digital Wallet capable of cryptocurrency and NFT transactions through supported bridge extensions or other Payment Processors approved by 404. Connecting your Digital Wallet through the Service is required for certain functionalities, including Token purchases. Once you initiate a Token purchase order, the Service forwards it to the appropriate extension, which executes the transaction on your behalf.
The project is structured into three main modules, each serving distinct functions within their framework for 3D content generation and validation:
Generation Module (generation): This module is central to creating 3D content and is compatible with miner neurons. It can also operate independently for development and testing purposes.
Neurons Module (neurons): This module includes the entry points for miner and validator neurons. Miners use RPC endpoints in the mining module to generate images, while validators retrieve and verify these images. It manages the execution of Bittensor subnet protocols and workflows.
Validation Module (validation): Dedicated to ensuring the quality and integrity of 3D content, this module operates alongside validator neurons or independently for comprehensive testing and quality assurance.
Hardware Requirements
Pending detailed benchmark results, their recommended setup aligns with Google Cloud’s a2-highgpu-1g specifications:
- GPU: NVIDIA A100 40GB
- CPU: 12 vCPUs
- RAM: 32GB – 85GB
- Storage: 200GB SSD
- Expected network traffic: Approximately 500GB per month under continuous operation, with a throughput of 0.2MB/s.
Minimal setup aligns with RunPod 1 x RTX 4090 specifications:
- GPU: NVIDIA 1 x RTX 4090
- CPU: 16 vCPUs
- RAM: 61 GB
Operating System Requirements
While their code is compatible with various operating systems, it has undergone rigorous testing primarily on Debian 11, Ubuntu 20, and Arch Linux. The Deep Learning VM Image provides the most stringent testing environment, equipped with pre-installed ML frameworks and essential tools for development.
Subnet Team Information