Can Makery.AI Create a New Marketplace of AI-Based Content Built Only on Authorized/Licensed Training Data?

Above: Real-life model/influencer Kate Purt, generated by Makery.AI creator vAIsual from a few hundred images of Kate — with her full and prior consent

Makery.AI is an upcoming platform for anyone excited by the power of generative AI, but struggles with the ethical conundrum of creating images and videos trained from the work of human artists without their consent or licensing — and relatedly, worry that an IP infringement lawsuit or regulatory body will eventually bring most of the broader generative AI industry crashing down around our heads.

Makery, by contrast, is a marketplace for creating new content built on a clean ecosystem of authorized training data — where creators can earn 50% of revenue from this work. Created by vAIsual, a startup already providing AI-based technology for Adobe, Google, and other companies, Makery’s trained models are only built on legally acquired content. In fact, some pilot artists/creators are already using it to generate new images built on their own data (but more on that later).

If you’re wondering if there’s a metaverse/virtual world tie-in to all this, you’re right: vAIsual COO Keren Flavell was the executive producer of SLCN.TV/Treet.TV, a broadcast-type service for Second Life, back in that day. So she has a strong background on the power of leveraging user-generated content:

“We are looking for people who want to experiment with creating their own AI models that can be licensed by others,” as she puts it. “So all those Second Life shop creators could do something similar — sell their creative output. Now converting SL assets is unlikely to work technically, but the designs (if they are original) could be created into models.

“For example the builder of historic battlefields [in a virtual world] might create a LoRA for Medieval War Lords and then someone can license that to generate still or video generative output.” (LoRA, for low-ranked adaption, is a kind of AI training model.) “Ad people might be customers, game designers etc. They can also make a LoRA of themselves, or their avatar if they have high quality images or video to train on.”

And yes, take your avatar into the real world, if you want: “You create a model that represents thousands of iterations. You can create your own avatar that you own.”

Assuming, of course, Makery.AI can gain traction against giants backed by Microsoft, Meta, and other incumbents:

“The high costs of training mean only the biggies have the money to train the models,” as she puts it, “and they are opting to exercise their god-given right (thank you America) to train on whatever the hell they want… for now at least. So all the glorious flooding of AI-generated content is coming from those very stained services.”

Flavell likens this moment to the time some twenty years ago when movies and music were being rampantly pirated online. The opportunity now is for generative AI to go legit. Instead, the leading AI companies are using their money and computing advantage to try and generate an insurmountable lead by building a massive base of training data, much or most of it unlicensed/unauthorized.

“There needs to be first of all regulation so it becomes a little scary to download the latest blockbuster,” as she puts it. “And the [legit] offering to be so good in the paid service, why would we bother with bittorrenting? We are in that phase right now.”

In this case, however, it’s not just about consumers who want to download movies and music for free:

“Commercial industry stands to gain the most from AI, genuinely reducing costs of production, notes Keren “However, right now it's suicide to use the standard services to produce commercial output. Waaaaaaay too risky for their clients.”

The deep ambiguity around copyright/IP rights means professional use of platforms like Midjourney scares professional content creators away. (Indeed, brands recently forbade ad agencies from using AI in their marketing campaigns.) Flevel estimates that major companies already spend millions in legal fees a year dealing with IP infringement disputes, even before AI-based IP rears its head.

But indie artists and creators can also benefit from Makery.AI, even earning income by helping contribute to the training database, “[both] the direct dataset for the LoRA as well as potentially in commercial datasets we sell to customers.”

Artist Anthony Breslin, for instance — currently wracked by cancer — is working with Makery.AI to train the model based on his own work to generate new, "Breslinator" imagery (above). And then any images sold, his estate gets a cut, even after he dies — a portion of which will be donated to a children's hospital.

And given all the Second Life content creators already experimenting with AI, I expect many would be interested in experimenting with Makery.

Keren agrees.

"If we look at the window that SL opened up, then generative AI is at the cusp of what it can do in terms of bending reality,” as she puts it here. “It’s a fundamentally different approach to creativity — it's generative, there's no folders and files when you're dynamically creating."

Expected to launch in coming months, you can sign up for early access to Makery.AI here. In fact, the first 50 people to mention New World Notes will get three months free use of the tool.