Training AI on Your NFTs

How to use TITLES new tool to train AI on your NFTs!
William M. Peaster William M. Peaster May 2, 20243 min read
Training AI on Your NFTs

Last August, I introduced Metaversal readers to TITLES, an NFT remixing platform with an in-built attribution and payment splits system. 

After its initial app + protocol combo release, the TITLES team stepped back and realized that the issues of attribution and creative ownership were particularly significant in the emerging AI arts scene. 

Recognizing this, they’ve since adapted their existing tools to better support AI creatives. Their first major step forward here is a functionality they unveiled today, the ability for artists to train their own custom AI models on NFTs they’ve made before.

The idea? Become a leading creator suite and minting hub for artist-trained AI models. 

I was fortunate enough to be treated to a preview of this new resource as TITLES founder Sören John was familiar with my collection of AI sketches, Stable Peasters, and invited me to test creating a custom model so anyone could make and mint their own Stable Ps. 

So I did! And I have to say, the process was buttery smooth and took only ~20 minutes. Here’s how the flow works in case you're interesting in trying it out for yourself: 

1. Go to and connect your wallet

2. In the “New Model” tab, select which of your NFT collections you want to train on—e.g. here I selected Stable Peasters

3. Select 5 to 20 NFTs to train your model on and finalize your model’s basic info—the more visually harmonious your selected inputs are, the smoother your model’s outputs will be, so curate thoughtfully

4. It will take up to 20 minutes for your model to finish training; you can track its progress in the “Your Models” tab

If you don’t have any NFT collections you’ve created of your own yet, but still want to play around with making AI remixes, try firing off some test prompts with my SPRY01 model, a.k.a. Stable Peasters Readily Yours V1, as seen above.

You can input as many prompts as you want without minting anything, but if you do come across an output you love, feel free to save the image or use the TITLES interface to mint the piece as an NFT on Base—just fill out the description and mint details as you please and press “Publish” to finish. 

What’s cool is that every time outputs are minted like this on TITLES, the original model creator (e.g. me, in the case of SPRY01) earns a small cut of ETH from the mint. This approach allows AI creatives to earn and directly benefit from attribution while still openly sharing their work with everyone else.

Looking ahead, keep an eye on TITLES as they enhance and expand upon this custom AI tool in new directions, e.g. allowing more than 20 NFT inputs and so forth. 

In the coming months, the team also notably plans to release a new AI-powered NFT publishing protocol that will offer artists innovative ways to engage with and profit from their models’ outputs. More on that later!

In the meantime, follow TITLES on Twitter or Warpcast to stay current with the project, and if you do make your own AI model or publish any of your own Stable Ps, don’t hesitate to hit me up—I’d love to see what you come up with and would love to mint one of your editions as a keepsake!

William M. Peaster

Written by William M. Peaster

655 Articles View all      

William M. Peaster, Senior Writer, has been with Bankless since January 2021. Immersed in Ethereum since 2017, he writes the Metaversal newsletter on the onchain frontier, covering everything from AI projects to crypto games, as the team’s lead NFT analyst. With a background in creative writing, he writes fiction and publishes art on Ethereum in his free time. He lives in Washington.

No Responses