MFL, previously named the Metaverse Football League, is using generative AI to reinvent the looks of the players in its soccer game.
For 180,000 players in the game, GenAI can now give them a new look to replace the previous stylized and cartoonish characters.
MFL was founded in 2021 in the midst of the metaverse and blockchain hype cycles. It started out with simple cartoon characters for its soccer players. But in the course of just a few days (and months of preparation), MFL re-generated every player with a more realistic look, said Mathurin Blouin, CEO of MFL, in an interview with GamesBeat.
“We didn’t do it to chase a trend. We did it because our universe needed it,” Blouin said.
Until now, MFL players were stylized and cartoonish. It served its purpose — but it also limited how real the game could feel. This overhaul was about immersion.
“We wanted a world that feels lived-in, with players that could believably exist in real football culture. Photorealism isn’t just about visuals — it changes perception. It makes MFL feel professional. It sharpens first impressions,” Blouin said. “And crucially, it opens the door to a future of dynamic storytelling and customization: custom jerseys, player aging, AI-generated video content. This opens up a world where our players will someday be able to give post-match interviews, have their own social media accounts, appear in ads or interact directly with their in-game agents and managers.”
Blouin clarified that MFL avoids Web3 tokens to prevent speculation, focusing on digital ownership of NFTs for monetization, and targeting mainstream adoption through a hybrid Web2/Web3 approach. Blouin highlighted MFL’s three-year operational longevity and projected a revenue target of $2 million by the end of 2025.
Behind the scenes, the company built a sprawling pipeline of over 400 interconnected processing units (or “nodes”) – each fulfilling a specific role in the system, like individual gears in a finely tuned machine.
The aim was to generate characters that look real, fast, and at scale. The entire pipeline was built using open source tools. No third-party black boxes. No rented AI APIs. Every part of the system is ours – from prompt generation to post-processing. That gives us full control, total flexibility, and zero dependency on proprietary models, Blouin said.
“Every image passed through multiple stages: It started with a prompt engine. A smart script that randomized player traits like their body type, hairstyle, expression, and more, to build a believable, varied universe from structured randomness,” he said.
These prompts were fed to over 100 high-end GPU pods, which generated base portraits with anatomical accuracy and regional diversity.
From there, a facial refinement pass gave each player a bit more soul – polishing expressions, fixing inconsistencies, and anchoring the emotional realism of each image. And then came the kits (i.e. the jerseys they wear).
“We built our own in-house 3D kit generation tool, which allowed us to design hundreds of club shirts using custom patterns, textures, and fictional brand logos. Those kits were passed back into the pipeline, and applied in a multi-step process: First, we used a segmentation model to detect the original shirt in the image, and then an advanced outpainting workflow to perform the clothing change,” Blouin said.
Logos – small details are the first thing to break down in AI generations – were enhanced in a separate detail pass. And a second queue of GPU pods ran background removal in parallel, outputting transparent images ready for the MFL platform.
Final quality control was handled in two layers: an automated check for obvious errors, followed by a human review team working inside a custom moderation panel we built. Rejected images were auto-regenerated. No manual fixes, no bottlenecks.
The Result
On the surface, it means MFL is going from cartoonish to cinematic. But in reality, it means so much more than that.
“We re-rendered 180,000 footballers, each one linked to their real in-game identity, fully isolated on transparent backgrounds, and ready to be used not just in the game UI – but in video, audio, social, and any other format we dream up in the future,” Blouin said.
The result is a world that truly reflects the essence of MFL: one that feels alive, like a breathing, dynamic parallel universe. One with lore. With identity. With continuity. And a system flexible enough to let us evolve every one of those players season after season – from teenage prodigy to retiring legend.
Blouin said the results of the improved engagement is also helping it go toe-to-toe with some of the biggest games in Web3 gaming. And he believes his smaller company can outperform larger rivals thanks to that engagement.
He believes it’s a rare example of a fully original Web3 game not only working, but scaling with true real creativity in their rollouts with technical depth.
The company started in late 2021, in the midst of the Web3 hype. But the company bootstrapped itself and did not raise external funding. It sold player and club NFTs in packs from the beginning. It built the ecosystem of 180,000 players who keep coming back. The average revenue per user is around $270, contributing to profitability and low operational costs.
The game is structured around six-week seasons, mimicking real football leagues with promotions and rewards, and players engage in managing clubs, recruiting players, and negotiating with other users.
Over time, MFL is aiming for mainstream adoption by targeting Web2 users and focusing on creating an enjoyable, emotion-driven experience rather than a speculative ones. The growth strategy will focus on increasing visibility, leveraging organic referrals and implementing paid user acquisition through influencers and future marketing campaigns.
Despite “Metaverse” in its full name (Metaverse Football League), Blouin noted that the company primarily uses the acronym MFL, similar to NFL or NBA, to avoid the current negative public perception of the term “metaverse.” He believes the metaverse concept has future potential, but their strategy prioritizes utility and gameplay before fully immersive 3D realizations.