Video Block Explorer

The video block explorer enables users to visually analyze on-chain activities using LLM.

Users can enter any address (and prompt) and generate a short video (max 1 minute) long, which can be used to easily digest data. The video is ready within 20 seconds.

It creates an analysis of the transactions, NFT activities, or interactions with smart contracts and uses human-readable diagrams, dashboards (potentially with AI voice and NFT images), and readable texts to explain any interesting events or summaries. After the video generation, users can decide whether to mint the video as NFT or not, also paying a small amount to the service, thus enabling this service to continue.

I used NextJS as a Full-Stack Framework, then Supabase for databases. For each user request, on the backend, it connects to Celo’s API to retrieve different types of data, which are later fed to the LLM to analyze and output a Structured output, which are going to be used as an input for the video render framework. On frontend, it connects to the user's wallet, to show their generated videos, or mint the video. The video part is handled using Remotion. When a user first generates a video, the remote actually just shows the video-like ReactJS UI. That’s how it achieves the fast generation. This also enables us to not spend resources on actual video rendering, if the video is not perfect. NFT Smart Contract is written in Solidity with Hardhat.