Fractal, taking inspiration from Manifold Labs' Deterministic Verification Network, is utilizing deterministic verification to create a grid of decentralized nodes to perform inference for video generation. By incentivizing users to host text-to-video models, this subnet allows Fractal to develop a mechanism that optimizes how end-user inference requests are handled to minimize latency. Additionally, the subnet is incredibly gamification-resistant, as a random seed is generated for each inference request, and if the Verifier and Prover seeds do not match, the Prover will be penalized.

GitHub

HW Requirements