GenHub | Devpost

Model Generation
Image Generation
Video Generation
Fetching all models
Fetching data of models

Inspiration

Online platforms such as TikTok or Instagram have their shops growing rapidly. These online platforms are capable of creating an online presence of a user which can influence thousands. If an influencer promotes a product, thousands of users get influenced and biased in their choices while buying. This platform provides a way to generate AI characters using generative AI which can be used for creating generative advertisements for products with these characters.

What it does

GenHub uses AI to generate real-people-like characters with prompt generation using dalle. These models are minted on-chain as token-bound accounts which are capable of holding their assets. Image and Video generation tools use these generated characters and make content around these as soul-bound NFTs which stay as holdings of earlier created token-bound accounts and cannot be moved to other accounts. This content can be fine-tuned and improved as per the taste of the user and then can be used for posting online to create an online presence for these ai-generated characters. These characters-based token-bound accounts with all their generated content can later be listed on the marketplace to be sold to other users. TheGraph substreams are used to query all of the characters generated and content around generated by running parallel fetching and bringing on-chain data to our frontend (unfortunately substreams didn't support Polygon Mumbai subgraphs so the fetching can only be seen on the terminal and couldn't be integrated with frontend on analytics tab).

How we built it

For the front end, we used Next.js along with Tailwind. To generate different kinds of AI results like model images, advertisement images or videos, we used a Python backend with Flask APIs. All the models generated are minted as token-bound accounts on the chain where the model image API is called from inside the smart contract using Chainlink Functions. For the video generation API, it wasn't possible to deploy it, so we spun up the API on our local host. All the content - images or videos generated by a model are stored as NFTs under that model's token-bound account.

Challenges we ran into

Video generation requires a lot of GPU power so we needed to buy the paid APIs for this purpose. Apart from this fine-tuning the Image generation API to perfectly generate content had been a difficult task as well. Ensuring a smooth connection between AI generation, NFT creation, and blockchain transactions required meticulous coding and testing.

Accomplishments that we're proud of

We were able to complete the project in the given time frame.

What we learned

We learned to use Javascript inside of smart contracts with Chainlink Functions, working with several sorts of AI APIs and fine-tuning their results according to our needs.

What's next for GenHub

We will improve and fine-tune the APIs to create more relevant and useful content. We will also be adding more features on content generation according to their taste for users.