SuperReferrals

SuperReferrals combines long-running agents with on-chain payments and deep user product integration to turn product/ref links into unstoppable marketing videos.

A framework that lets anyone sign up either as a store owner and set their own generative pricing and policies, ENS identity, and custom branding for their store, or as a user of an existing store and create generative marketing videos from their ref links.

Long-running agents are great. In fact, when using the latest models, which can handle really long contexts, longer-running agents can achieve some pretty impressive feats.

However, there is a problem: long-running agent tasks are expensive, consume a lot of tokens, and sometimes fail. Keeperhub helps us solve this problem and allows cross-chain payments.

This problem explicitly hurts the end user, who doesn’t want a refund in credits or to purchase any plan. They just want to create a video from their image list + metadata and get a refund if the task fails.

Transaction reverts are handled via Keeperhub. The store owner gets refunded in credits, while the user gets refunded in crypto.

Stores are powered on the backend by samsar-js, a model-agnostic API enabling the creation of marketing videos from ref links up to 3 minutes long.

Store owners can set their model preferences, policies, and whitelists, and customize their storefront branding, as well as meta customization such as specializing in creating marketing videos for electronic stores or anime-style video stores.

Users choose the store they want to interact with on the storefront generate video page, connect their wallet, add images, choose the model, enter a prompt, and make payments for their generation settings.

Once the long-running generation task finishes, the user can view, download, or share their rendered video from the generation page itself or from the INFT page.

Users can further set the pricing for “Edit rights” for their INFT. If another user wants to obtain edit rights, they can pay the price to purchase a deep clone of the original INFT. The purchase in this context is not a transfer, but the purchaser obtains a copy of the INFT on which they can do meta operations such as re-roll/switch scenes, audio, retranslate audio, switch outro, or CTA URL. All transfers and edit operations create new INFTs that can be audited on-chain.

Anyone can download and share the video anywhere, and because the CTA/ref link always points to the original author website and monetization is no longer tied only to link views from the original account, video shares by third parties help rather than hurt the original creator.

If someone wants to insert their own ref link/CTA, they will need to purchase the video INFT from the original creator, only if third-party purchase is enabled by the creator.

Fully customizable, fully auditable generative marketing videos as INFTs via cutting-edge long-running agent technology.

SuperReferrals is built on Next.js. It provides a framework for building generative video stores and also contains a sample crypto-native e-commerce store with a built-in partner referral link mechanism.

The video generation backend uses samsar-js for long-running generative video tasks, enabling stores to turn product/ref links, image lists, metadata, and prompts into editable marketing videos.

The crypto backend uses Solidity smart contracts with Hardhat for development, testing, and deployment. Payments and transactions are made on Sepolia/Base, gas is paid on the 0G network, and routing is handled via KeeperHub.

ENS is used for store owner identity and branded storefront resolution.

Uniswap is used for payment quote discovery, best-price routing, and swap transaction creation when payment and settlement tokens differ.

KeeperHub handles transaction reverts and refund routing, so store owners can be refunded in credits while end users can be refunded directly in crypto when long-running generation tasks fail.

Together, these pieces create a crypto-native framework for fully customizable and fully auditable generative marketing videos as INFTs.