Iconic Children’s Show ‘Pocoyo’ Has A New Owner — And A New AI-Powered Production Pipeline

Pocoyo, the 3d-animated boy with the blue beanie and button eyes, is a courageous soul. The mostly mute, adventurous little hero — creation of Guillermo García Carsí, Colman López, Luis Gallego, and David Cantolla — occupies a unique place not only in Pocoyo World, the usually all-white environment where he lives with his colorful mostly-animal friends, but in the hearts of generations who grew up with the characters over more than 20 years.

Things changed for Pocoyo in 2023 when his then-owner, production company Zinkia Entertainment in Spain, sold all rights to their beloved characters to Animaj in France. Animaj has since emerged, under the guidance of entrepreneur Sixte de Vauplane — creator of France’s Nestor food delivery app — as a force in animation software that is intrinsically linked to Pocoyo’s next stage of evolution. De Vauplane co-founded the company with Gregory Dray, who was formerly the managing director, international, for Youtube Kids & EMEA Kids, Family & Education.

Late 2023, García Carsí, the key creative force behind Pocoyo, joined Vauplane’s team as creative director and is now midway through shaping season five of Pocoyo’s new adventures. The Animaj toolset is also changing with the introduction of Sketch-to-Motion and other workflows that are bridging gaps that some have feared to tread, using artificial intelligence with the caveat of artistic control, offering interfaces designed to cater to graphically-talented, seasoned animators.

Cartoon Brew spoke with de Vauplane to discuss Animaj’s philosophy toward technology, his collaboration with García Carsí, and upcoming plans for their eternally youthful animated boy.

Cartoon Brew: Did your interests in AI stem from Nestor?

Sixte de Vauplane: It’s completely new. I’m a tech guy, so I was focused on technology. During the genesis of Nestor, I was using Youtube as a way to watch movies with my three boys, as many parents do. After selling my company, I wanted to create another. I decided to focus on kids’ media. As a parent, I knew we were using Youtube and digital platforms, all day. We don’t consume content in the same way as before. And I believed there was a need to create a new model in the industry. I decided to create the next-generation kids’ media company that could build global kids’ franchises, but with a new approach, using generative AI. I wanted to design a new way of producing and distributing content.

How did you develop that?

I was fond of Nvidia’s Audio2face methodology. That was more focused on the gaming industry, and I wondered why couldn’t we apply gen-AI models to the animation industry. After researching how it worked, I understood that the barriers between gaming industry and animation were fading away as the technology was improving. My conviction at that time, in early 2022, was that we were not so far away from a disruption that would reshuffle the industry. Technology was maturing in a way that could make the animation production much more scalable. That was the turning point.

Animaj’s tools appear to have an understanding of the physical and artistic side of animation. Sketch-to-Motion, for instance, looks like it complements the traditional animator skillset. Was that a goal, and can you draw?

I’m very bad at drawing. (laughs) I used to [draw] when I was young, but I’m too much on the engineering path [now]. I know people are freaking out about AI replacing the creative artist. For me, that is not a good debate, because I’m not sure that gen AI is a good creator. Gen AI is good when it serves creative artists. The way we design every AI tool is to empower the artists; not to not replace them. We understand that creators are used to sketching, and it’s easier for them to draw their creative vision. For them, there is nothing more frustrating than having to wait weeks, maybe months from the sketch to animation, and the final rounded content. Creators know what they have in mind. And they want to have a way to bring it to life faster, so they can iterate further, so they can keep increasing quality. The way we’ve been working in Sketch-to-Motion is to remove the tedious, time-consuming parts of the animators’ job, so they can concentrate on the creative vision and the capacity to iterate. We know that the number of iterations is a key factor to elevate quality. That’s typically what our Sketch-to-Motion can deliver.

In creating iterations, how does Sketch-to-Motion learn the animator’s style?

Without getting too technical, we train our model on our own proprietary storyboards, sketches, and 3d poses, and then we’ve been using data augmentation techniques to generate a range of variations from those original drawings. That’s how the model can ‘understand’ Pocoyo and interpret, as much as possible, the storyboard artists’ drawings.

The Animaj website mentions ‘convolutional neural networks’ — is that what you’re describing?

It is a ‘RAISE’ architecture (‘Reasoning and Acting through Scratchpad and Examples’). That sounds barbaric! But that is Sketch-to-Motion’s underlying technical architecture.

To demystify that a bit, when the artist does ‘Motion In-betweening’ from drawn keyframes, can you talk us through how that works, creating Pocoyo’s snappy little moves?

Sketch-to-Pose and Motion In-betweening are two sides of the same coin. From the sketch and from the animatics, we generate the key pose, and then we generate all the inbetween poses by running a proprietary interpolation model. The difference [between this and] the Maya interpolation model, or the Unreal Engine interpolation model, is that we’ve been training our model for the specificity of our own characters. That’s key to our approach.

We’ve been building a framework where we can create, in a scaleable way, one model per character. And it takes approximately 24 hours to set up a new model for a new character. With Pocoyo, it is so important that he is pose-to-pose. The interpolation model from Maya would create a very smooth [animation] curve, from point A to point B. But if Pocoyo is animated smoothly, he’s not Pocoyo anymore. That’s why we’ve been training our model to replicate Pocoyo’s exact animation style.

It is super important to build our proprietary training data sets specific to every character. From a technical and philosophical standpoint, it is the exact opposite of the commercial gen AI technology, like Sora, Pika Labs, or Luma. Those large models have been trained on stealing massive, third-party data sets, often scraped without specificity, which raises ethical and IP issues. They aren’t tailored to any one animation style or character. If you tried to use Sora to generate Pocoyo, it would not capture the distinctiveness of his personality, or the uniqueness of his animation.

How instant is the creation of a 3d sketch in Sketch-to-Motion?

Motion In-between is just now [becoming] real-time. In the demo, a few months ago, we had one hour of compute-time. It is now less than a second, or 0.5 seconds of loading time. Sketch-to-Pose is real-time. You click a button, and you get the pose in less than three-quarters of a second. And the output of our AI model is a 3d editable file. That means, even if it is not perfect, the animator can tweak it, and iterate from there.

Let’s talk about acquiring Pocoyo. He was obviously a major coup for Animaj, as he is so loved. How are you making sure, working with Guillaume, that you are keeping true to the spirit of that character?

We believe that to step up the creative game on digital platforms, including Youtube, it is our responsibility to make sure that the most beloved and high-quality IPs can be there. That’s why we acquired Pocoyo. And we hope to announce, in two to three months, two other major acquisitions that will be similar to Pocoyo in terms of the quality, distinctiveness, and uniqueness. But Pocoyo was important, because he is a big part of Spanish culture.

Talk about how your relationship started with Guillermo?

It was pretty funny. The first time I reached out to him, I sent him a Linkedin message, and I was pretty sure that, at first, he wouldn’t answer. But he did it straight away. And then, when I first met him at the bar in Madrid, I expected he might be into other things because it had been 10 years since he left Pocoyo. After a few hours’ discussion, I knew that Pocoyo was still there in his heart. He had a clear vision of how to refresh Pocoyo. We discussed that for a few weeks, and we realized that we were fully aligned in our vision. We wanted to refresh Pocoyo for the next 20 years. That’s when I asked Guillermo to come on board, to direct and lead the creative vision of Pocoyo.

How are you developing that vision?

We have a five year roadmap. We want Pocoyo to be an icon for the next generation, and that starts with content. We’re releasing digital-first content to engage with the audience weekly, on different, shorter, more dynamic formats. From a creative standpoint, Guillermo is at work on Pocoyo’s season five, whose second part is going to be released in the coming weeks. It will be clearly different from the others. Pocoyo is finally speaking, after 20 years. Pocoyo has a little sister who is creating immense chaos in his life, keeping him focused on adventure. It’s a bold, creative risk. But we believe that if we want to take Pocoyo to the next level, we need to take some risk, and with Guillermo leading we make sure that we don’t compromise what makes Pocoyo so unique. It’s what makes Pocoyo so charming.