How Freepik Transformed From a Stock Image Platform Into a Generative AI Powerhouse – Crypto News

Fifteen years ago, Freepik was just another stock image provider, helping designers find the right visuals for their projects. Today, it’s a completely different beast—a generative AI hub attracting over 60 million visitors a month.

The shift wasn’t accidental. It was the result of a company willing to rethink its purpose and move beyond static images to something much bigger: a fully AI-powered creative ecosystem.

Sitting in his living room , Freepik CEO Joaquín Cuenca Abela spoke with Decrypt about the state of the AI industry and how his company harnessed the opportunity generative AI brought for digital artists.

“When generative AI appeared, we saw we could expand our mission,” he told Decrypt. “We were no longer limited to helping designers with pre-made content. Instead, we could adapt to what they needed and create something unique for each person.”

The move paid off. Freepik is now a one-stop shop for AI-powered image and video generation, upscaling, animation, and more. Cuenca Abela put it simply: “We just want to give creatives more control.”

From blank page to AI engine

Freepik started with a simple premise: eliminate the frustration of staring at a blank page. Before AI, the platform provided millions of stock images and templates that creatives could use as starting points.

“The slowest, most painful part of the creative process was starting from zero,” Cuenca Abela said. “We helped eliminate that barrier by giving designers millions of images they could begin creating with.”

Now, with generative AI, Freepik doesn’t just offer a library of content—the firm creates it on demand.“A photographer wasn’t a traditional Freepik user,” said Cuenca Abela. “They already had their images. But now, with our upscaler Magnific, they can enhance them in ways they never could before.”

The shift has broadened its audience beyond graphic designers. Photographers use it to enhance, tweak, and upscale images. Filmmakers experiment with AI-generated visuals, architects and interior designers build concepts in ways that were once time-consuming and expensive, and the average Joe uses it to generate beautiful waifus—because, of course, it’s AI we’re talking about.

Not just another image generator

In the crowded AI space, Freepik is focusing on workflow integration. Most AI tools specialize in one thing, be it image generation, video creation, or upscaling. Freepik connects them all, acting like a hub that integrates different open- and closed-source generative AI tools in one place.

Among other services, the company’s AI suite includes:

• Image generation with models like Mystic, Flux, Ideogram, and Google Imagen
• Custom LoRA training for consistent character and style generation
• Video generation using seven different models, including Google’s V2, Hunyuan, Luma, Kling, Hailuo, and Minimax
• Editing tools for inpainting, outpainting, filters and seamless image expansion
• Audio generation, including music, voice-over, and sound effects
• SVG conversion capabilities for vector-based assets.

One of the biggest success stories has been Magnific, Freepik’s AI-powered upscaler. It went viral for its ability to enhance image details without distorting them—something even top-tier AI models have struggled with.

Then, Mystic was the icing on the cake, providing results that were capable of competing against state-of-the-art models like Ideogram or MirJourney. Mystic is actually a workflow, using Flux as a core model with a lot of tweaks behind the scenes.

That said, a lot of experts and enthusiasts have tried to mimic Freepik’s secret sauce—which relies on open-source models—with mixed results. Freepik has a way to provide quality results, consistently, with the best user interface possible, which is what customers are paying for.

“People sometimes underestimate the difference between a good product and an excellent product,” Cuenca Abela said. “The last 10% takes 90% of the effort. That’s why many tried replicating Magnific, but couldn’t quite nail it.”

The copyright debate

With AI-generated content comes controversy. Many artists argue that AI developers unfairly train their models using copyrighted works without permission. Cuenca Abela doesn’t dismiss their concerns, but said he sees the issue differently.

“If you required permission from every individual creator to train an AI model, those models simply couldn’t exist,” he said. “It’d be like asking permission to index every single web page before launching Google.”

He acknowledges the tension.

“This in the short term damages the artist—using something created by the artist. For the [affected] artists, this is a situation of profound injustice,” he told Decrypt, recognizing that such advancement forced them to evolve as business. “Something similar happened to us. When (AI) emerged, our business suddenly brought less value to the table. We had to adapt.”

That said, he argues that AI-generated images aren’t direct copies. “The complaint traditional artists usually have is that their images were used without permission, which is totally true,” he admitted. “But the counterpoint is that the images these models produce aren’t copies. If a person had made them, there wouldn’t be a claim of copyright infringement.”

The core of the debate is basically the trade-off between creative control and technological progress. Cuenca Abela believes society will ultimately favor AI’s benefits—just as it did when similar debates surrounded photography killing painting, digital art killing traditional art, or internet search engines killing encyclopedias.

“As a society, we’ll need to balance things and make a decision. If permission from the creator is required to train a model, generative models for text and images won’t exist.” he said. “[If that happens, then] society loses all the progress that text models provide. They can help us find vaccines, medicines; the scientific advancements they can bring are tremendous. All that progress is lost.”

Cuenca also sees AI as a tool for self expression. He doesn’t differentiate AI artists from artists.

“There’s no difference. It’s a tool. AI is a means to express what you want, and art is the expression of what’s inside you, what happened to you, your life experiences—well, you can do that with AI, with paintings, with photos… It will depend on the artist,” he told Decrypt.

“For me, it’s totally art and it’s legitimate. I don’t have an ethical problem with that.”

Open models vs. closed systems

There used to be consensus about closed-source being the go-to option for end users, with models and technologies usually being more user-friendly and providing a better-quality experience than open options. However, things have drastically changed over the years.

Stable Diffusion revolutionized AI art, Llama was key to bringing local text generation to the masses, and more recently, DeepSeek R1 reignited the debate about closed-source AI companies overcharging for their models.

However, some users still prefer closed-source options. Cuenca Abela has strong opinions about the AI industry’s future, particularly in the battle between open-source and proprietary models.

“In terms of code, state of the art open-source is at the same level as proprietary models,” he said. “The biggest difference is training time and dataset curation, a longer post-training phase, a bit better tagging, etc. But as for technical level, I don’t see much of a gap.”

While proprietary models like MidJourney and Ideogram get more refinement, Cuenca sees open-source alternatives closing the gap quickly. He points to Flux as an example: “It might be a tiny step behind the best closed models, but not two steps. And because it’s open, the community fine-tunes and builds on it, sometimes surpassing the closed versions.”

For Freepik, variety and flexibility are the priorities. “Someone who knows how to use Freepik will get better quality than MidJourney,” Cuenca Abela said. “If you need photorealism, we have Google Imagen. If you need artistic text generation, use Ideogram. If you need character consistency, train a LoRA. No single model is the answer to everything.”

In other words, there’s no jack of all trades in AI. And the versatility of choosing open and closed-source models on demand is extremely important to get the granularity required for the perfect work of art—one that truly resembles what the user has in mind.

Freepik’s AI video bet

Recently, Freepik has doubled down on AI-powered video tools. The company integrated Google’s Veo 2, which dramatically improves video generation quality.

“Before Veo 2, you had to generate 10 or 20 videos to get one that worked,” Cuenca Abela notes. “Now, with Veo 2, you get a good result every other try.”

But the real game-changer for video artists will be an upcoming AI video editor, he said. Instead of just generating short clips, users will soon be able to assemble full videos entirely inside of Freepik.

“Today you can only make video clips—only generate small clips of 2 seconds, 3 seconds, 8 seconds. We are working on something that allows people to edit them on the page itself, add audio, and do the whole composition so that you end up with your clip,” Cuenca Abela told Decrypt.

“The goal is to make Freepik the creative hub where you don’t need to leave the platform to finish a project,” he said.

The future of AI: Opportunity or concern?

Are we close to artificial general intelligence, or AGI? Will machines replace us? Cuenca Abela sees AI’s rapid development as both exciting and unsettling.

“[AGI] feels close now—much closer than anyone expected just a few years ago,” he admitted. “We went from people dismissing AI as a toy to machines that can think.”

There is not really consensus about what exactly constitutes AGI, but it can be broadly conceived as a type of artificial intelligence that can understand, learn, and apply knowledge across basically any field at a human-like level or beyond, being capable of adapting to new problems. We are currently in a state of “narrow AI,” with models that excel at some things but underperform at others.

That shift, he argued, raises big existential questions. “Machines can be paused, restarted, or copied. Humans can’t. Those differences matter,” he said. “It means we’ll always have a unique place alongside technology.”

While some fear AI replacing human creativity, Cuenca Abela remains optimistic about its potential. “I think this will cause a very profound and strong acceleration that feels a bit overwhelming. We don’t know what we’ll be able to achieve in the future.”

A bit more down to earth, he thinks the immediate future may bring us more tools that help machines understand exactly what the user wants, being much more accurate and providing higher-quality results. And Freepik’s new philosophy appears to point toward that direction, becoming a hub in which artists can find everything they need to turn an AI generation into their own imagined work of art.

“That’s our mission: Helping people make great designs to express the power of their ideas,” said Cuenca Abela. “AI for us is just a tool—but it’s how people interact with AI that matters.”

Edited by Andrew Hayward