Online graphic design platform, Freepic on Tuesday unveiled a new “open” AI image model, which the company says that was specially commercially licensed, trained on “safe-work” images.
Model called F lightAbout 10 billion parameters are included – parameters are internal components that make models. F Light was developed in partnership with Fin Startup Fal.Ai and according to Freapic, was trained using 64 Nvidia H100 GPU during two months.
F lights are involved in a small and mounting collection of a trained generic AI model on licensed data.
We have been secretly working on it for months! Finally it feels good to share it!
Link:
• Regular version: more forecast and quick-believed, but less artistic:
• Texture version: more chaotic and error-prone, but better texture and … pic.twitter.com/gx5mipye8o
– Javi Lopez ⛩ (@Javilopen) 29 April, 2025
The generative AI AI is at the center of copyright cases against companies, including Openai and Midjourney. It is often developed using mass materials – including copyright materials – from public sources around the web. Most companies that develop these models argue Proper use The owners mold their practice of using copyright data for training without compensating. Many creators and IP right holders are disagreeed.
Freepic has provided two tastes of F Light, both standard and texture were trained on the internal data set of about 80 million images. The standard is more predicated and “quick-fetful”, while the texture is more “error-confident”, but according to the company, provides better textures and creative compositions.
Here is an image from the standard model, which originated with “a person standing in front of a sunset in the royal environment”.
Freepik does not claim that F Lite Midjourney’s V7, Black Forest Labs’ fluxes family, or others produce better images than the generator like family, or others. The goal was to openly provide a model so that the developers can tailor and improve it according to the company.
It is being said, running F Light is not an easy achievement. The model requires GPU with at least 24GB Vram.
Other companies developing media-generating models on licensed data include Adobe, BRIA, Getty Images, Moonvalley and Shutterstock. Depending on how AI copyright cases move, the market may grow rapidly.
