This project was another great learning experience. It's incredibly fast to build full stories with #genAI. And it's certainly way cheaper. But getting all of the details right is really the trick. That's where we focus: consistent characters and scenes, controllable motion, and script understanding

Like we've written about before, we think that the best way to *really* evaluate any creative generative media technology is not benchmarks, but rather to just tell stories. 2ndsetai.substack.com/p/the-only-b...

Lots more details around how we built this and what we found in the post on #Substack.

#genAI images don't have to be just copies of copyrighted IP. These same technologies can be used as tools for creatives to do new work. And part of getting creatives comfortable with these tools involves working harder to prevent infringement.

Our hope is for any IP rights holder to be able to understand just how often their stuff is being replicated. And see evidence of when the minority of providers are actually trying to keep their IP safe.

This piece is just a preview of what's to come. Make no mistake this work is really, really hard. Most #genAI solutions today do very poorly when put to the real world tests of consistency, controllability, and grounded character accuracy. We do lots of #AIresearch to advance these capabilities.

We really believe that this is the future of #anime. The world of #manga and #lightnovels already has many more stories than anime, due to cost and speed considerations. We see a world where more of those stories get fully realized in #genAI anime.

So, you get your characters, in your styles, telling the stories that build from worlds you've already created. This is way beyond, "I made my cat into a cartoon," type consumer use cases. This is a professional grade platform for real studios and publishers.

Of course, the really challenging part is getting it to follow *your* artistic intent, telling the story you want to tell, how you want to tell it. And that's where our unique technology at 2nd Set AI really shines. We built this tech from the ground up to actually serve people who have existing IP.

This piece was developed with our platform underpinning every step of the process from concept, to script, to shot breakdown, images, video, dialogue, sound effects, music, etc. You really can do it all with #genAI these days.

This is part of our larger conversation with the field of publishers and studios at #AnimeNYC this week. What we're consistently hearing is that the magic of this technology allows more organizations to consider #animation than ever would have before.

And this line of thought can go so much further. Learning new IP that you're actively creating can allow #genAI models to actually play an active and supporting role in content development.

Of course, my interest is in how can you do first party generation right. That's why we built this technology. We can detect potentially infringing imagery and thus keep it out of your studios generative workflows.

Perhaps more interestingly, we found that some properties were too new to have been learned by the models. The Apothecary Diaries and Blue Eyed Samurai were both too fresh to have been scraped up yet.

As usual, we found that there was a ton of IP coming out of commonly available generative image models. Major properties across geographies, media types, genders, and genres were all represented.

If you want to read the more narrative intro to this work, check out our Substack for a deeper dive: 2ndsetai.substack.com/p/introducin...