Artificial Influence

Uncanny AI Ambassadors for Unlimited UGC & Ads

Comment

Inspiration

Every day my part-time job is scrolling through reels for a good 1-4 hours a day, and before you judge my time i'm working on it ebcause i used to put in full time shifts. In the recent two months, I've seen a crazy rise in AI-generated influencers and influencer content. I'm talking about completely AI profiles that are getting hundreds of thousands or sometimes even millions of followers (a simple reel with 160m views in 2 days is what woke me up to this idea). Most people, especially those that dont keep up with the latest AI advances, have absolutely no idea that these are fully AI influencers. So I want to capitalize on this opportunity that state-of-the-art model that cross the uncanny valley edge bring. This new rising market has only a few platform that can do this and that makes the process easier than ever. Removing the need to research the best model, instead wrapping it up into an easy-to-use tool for the user or business that wants to create content

What it does

Artificial Influence is an all-in-one AI creative platform that lets anyone generate, animate, and lip-sync hyper-realistic digital influencers. Create professional UGC, A/B test ad creatives, and scale content without expensive shoots or unreliable talent.

How we built it

For this project, I decided to go with the tech stack I'm highly familiar with. I used Next.js app router for both the front-end and server-side routes for all the backends api route because a monorepo, allowed me to develop faster and have greater oversee of the code of all developing as one person over 36 hours.

For my database and storage, I use supabase with appropriate role-level security restricting access to all data at the object level, maintaining privacy best practices, and allowing me to easily scale in the future.

For styling, I decided to use my favorite component library as always: ShadCN UI with their new create customization to customize certain brand elements from the original ShadCN such as the icon kit, corner radius, average and other things. Youll notice this 3D depth effect everywhere in the site with lots of shadows in both light and dark mode. This was something new I tried insipred by a youtube video (https://www.youtube.com/watch?v=wcZ6jSlZqDc) customizing ShadCN base componenet to make the effect easy to use everywhere for consistency.

Challenges we ran into

I honestly ran into a lot of challenges over the 36 hours I was building it. But the biggest ones would definitely have to be:

  1. Customizing ShadCN base components with a custom depth effect to change their drop shadow and other things, like inset highlighting.
  2. Weird Google sign-in auth bug, which I still can't tell where it originated from but cursor fixed it.

  3. Finding the best models. Model selection means a lot, and to achieve the best results, I had to do a lot of testing and burn a lot of Replicate API credits. I tend to keep up with the latest large language models, but I don't do the same for the generative AI models for other media types because its too much.

  4. There was definitely more LOL

Accomplishments that we're proud of

I'm incredibly proud of the user interface and user experience I was able to achieve in such a short time frame. I focused on simplicity and positive transfer from existing AI chat sites and AI generator sites by implementing both the row (Chatbot) and column (Image generator) layout even making my site mobile responsive. While I'm with this, I'm very proud of how many tools and integrations I was capable of building within 36 hours. These integrations include image generation with all of the state-of-the-art models, including nanoBanana Pro, SeeDream 4.5, GPT Image 1.5, and more. Beyond this, I was also able to make simple, easy to use custom tools for motion copy using state-of-the-art motion transfer models like kling 2.6 and Veed Fabric 1.0 for lipsync.

What we learned

During this hackathon, I had to do a lot of research on the state-of-the-art generated AI models to host them in a scalable manner. To accomplish this, I decided to use the Replicate API, an AI infrastructure provider that allows you to run any model avalible. Built in alone in such a short time period I learned that composure and sleep were absolute necessities for achieving high levels of quantitative output in a time crunch. And keeping yourself low on sleep is the biggest productivity scam.

What's next for Artificial Influence

To make Artificial Influence a platform that can create truly life like videos. We're going to continue adding more easy-to-use, high-abstraction tools for creating these videos like motion control and lip sync studio, as well as always keeping up-to-date with state-of-the-art AI models, so that our quality only goes up with time. This project was based off real market trends of an inclusion AI influencers on Instagram and TikTok, immediately after the hackathon, we will add the strata units and post on Product Hunt twitter.com/ex to see if this project can be validated by the market, or if i miss judged the trend.

Built With

  • cursor
  • kling
  • nanobanana
  • next.js
  • replicate
  • shadcn
Share this project:

Updates