Freepik, the online graphics design platform, presented on Tuesday a new “Open” AI Image model that the company says it was trained exclusively in commercial images “Safe-For-Work”.
The model, called FlirtationContains about 10 billion parameters – parameters are the internal components that make up the model. F Lite was developed in collaboration with AI Startup Fal.Ai and trained using 64 GPUS NVIDIA H100 over two months, according to FreePik.
F Lite joins a small and increasing collection of AI genetic models trained in licensed data.
We have been working secretly for this for months! It’s good to share it in the end!
LINKS:
• Normal version: more predictable and fragile, but less artistic:
• Texture version: It is more chaotic and prone to errors, but it offers better textures and… pic.twitter.com/gx5mipye8o
– Javi Lopez ⛩️ (@javilopen) April 29 2025
Genetic AI is at the heart of copyright against AI companies, including Openai and Midjourney. It often develops using huge amounts of content – including copyright -protected content – from public sources around the tissue. Most companies that develop these models support fair use It deals with their practice of using copyright -protected data without offsetting the owners. Many creators and IP rights holders disagree.
Freepik has made two flavors from available, templates and textures, and both have been trained in an internal data set of about 80 million images. The standard is more predictable and “easy to use”, while the texture is more “prone to errors” but offers better textures and creative compositions, according to the company.
Following is a picture of the standard model created by the exhortation “a person standing in front of a sunset, in a magnificent environment”.
Freepik does not claim that F Lite produces images superior to top image generators such as Midjourney’s V7, Midjourney’s Flux family, or others. The goal was to make a model openly available so that developers could customize and improve the company, according to the company.
This is what the flute is executed is not an easy achievement. The model requires a GPU with at least 24GB VRAM.
Other companies that are developing models of media creation in authorized data include Adobe, Bria, Getty Images, Moonvalley and Shutterstock. Depending on how AI’s copyright lawsuits are moved, the market could exponentially grow.
