Created on 16th June 2025
ā¢
3D animated realistic icons are beautiful and aesthetic to use in different kind of applications. We can see the latest release of the Airbnb design are using these icons and had quickly becoming a popular trend.
š¤ If we want to create those icons using AI tools, what is the process?
It probably involve a user, have an icon in his mind, crafting a prompt for that, and feed it into an image generative model. And the output will be a static image icon. But there is more to do. The user probably need to do preprocessing or editing. And he also needs to think about how to animate it and write a script for this. Finally, feeding it into a video generative model, and produce the final animated icon.
š āāļø Wait... the pipeline is not optimum (problem)
There is too much manual work for the user. It is also time consuming if large amount of icons needed. It also involved lots of platform jumping to get things done.
This setup utilise Amazon Nova FMs with Amazon Bedrock.
š” Imagine we have 4 agents in our team. A planner agent, an illustrator agent, an interpreter agent, and an animator agent.
The pipeline starts from the planner agent, powered by Amazon Nova Micro, will analyse the user requirement and summarise what icon the user actually needs. Then, it will pass the information to an illustrator agent powered by openai image gen 1, which have a specific 3d icon design style system prompt, and will be generating that icon. This agent also equip with tools for the preprocessing steps. Next, the interpreter agent will look at the icon image output, as it is powered by the Amazon Nova Lite with vision capability. The agent will describe what it sees. It is super important as the animator powered by Amazon Nova Micro will write an animation script based on the description and start to animate, with the help of Runway gen 4 turbo.
šŖChallenge: We aimed to encapsulate our agentic pipeline into an API service, but several user experience hurdles emerged. We didnāt want users to deal with the friction of managing multiple API keys, purchasing credits, or creating and configuring accounts. Our goal was to make the process as seamless as possible with minimal setup.
āSolution: Uniicon addresses these challenges by integrating an x402-protocol powered paywall, enabling seamless monetization on Base. With this system, users simply make a generation request and complete a USDC paymentāno account setup or external API key management required. All payments are funneled into a revenue pool managed by a CDP wallet, which then automatically and transparently distributes earnings equally among the serviceās creators or owners.
šŖChallenge: Since Uniicon is designed as a lightweight framework, we avoid heavy storage dependencies. However, itās crucial that users are still able to re-access and download their generated icons anytime, from anywhereāwithout being reliant on the availability of our service.
āSolution: To enable decentralized and persistent access to generated icons, Pinataāleveraging IPFSāis the ideal solution for our architecture. Each time a user completes an icon generation, an AI agent automatically uploads the asset to Pinata, storing it on IPFS. Simultaneously, a reference token (NFT) containing the iconās metadata is minted on Base and sent to the userās wallet. This allows users to effortlessly access and download their icon by simply tapping into their token.
Uniicon has a MCP server, which allows you to connect external clientsāsuch as Claudeāto tackle complex tasks like UI design or web development. Moreover, Uniicon is agent framework agnostic, offering flexibility in building agentic system.
Tracks Applied (5)
Amazon Web Services
Pinata
Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.