{"id":61839,"date":"2025-04-30T02:58:08","date_gmt":"2025-04-30T02:58:08","guid":{"rendered":"https:\/\/www.europesays.com\/uk\/61839\/"},"modified":"2025-04-30T02:58:08","modified_gmt":"2025-04-30T02:58:08","slug":"freepik-releases-an-open-ai-image-generator-trained-on-licensed-data","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/uk\/61839\/","title":{"rendered":"Freepik releases an ‘open’ AI image generator trained on licensed data"},"content":{"rendered":"

Freepik, the online graphic design platform, unveiled a new \u201copen\u201d AI image model on Tuesday that the company says was trained exclusively on commercially licensed, \u201csafe-for-work\u201d images.<\/p>\n

The model, called F Lite<\/a>, contains around 10 billion parameters \u2014 parameters being the internal components that make up the model. F Lite was developed in partnership with AI startup Fal.ai and trained using 64 Nvidia H100 GPUs over the course of two months, according to Freepik.<\/p>\n

F Lite joins a small and growing collection of generative AI models trained on licensed data.<\/p>\n

\n

We\u2019ve been secretly working on this for months! It feels good to finally share it!<\/p>\n

LINKS:<\/p>\n

\u2022 Regular version: more predictable and prompt-faithful, but less artistic: https:\/\/t.co\/MyWsKer9Ir<\/a><\/p>\n

\u2022 Texture version: is more chaotic and error-prone, but delivers better textures and\u2026 pic.twitter.com\/GX5mIpYE8O<\/a><\/p>\n

\u2014 Javi Lopez \u26e9\ufe0f (@javilopen) April 29, 2025<\/a><\/p>\n<\/blockquote>\n

Generative AI is at the center of copyright lawsuits against AI companies, including\u00a0OpenAI<\/a>\u00a0and\u00a0Midjourney<\/a>. It\u2019s frequently developed using massive amounts of content \u2014 including copyrighted content \u2014 from public sources around the web. Most companies developing these models argue\u00a0fair use<\/a>\u00a0shields their practice of using copyrighted data for training without compensating the owners. Many creators and IP rights holders disagree.<\/p>\n

Freepik has made two flavors of F Lite available, standard and texture, both of which were trained on an internal dataset of around 80 million images. Standard is more predictable and \u201cprompt-faithful,\u201d while texture is more \u201cerror-prone\u201d but delivers better textures and creative compositions, according to the company.<\/p>\n

Here\u2019s an image from the standard model generated with the prompt \u201cA person standing in front of a sunset, in majestic surroundings.\u201d<\/p>\n

\"Freepik\"An AI-generated photo from Freepik\u2019s F Lite model.Image Credits:<\/strong>Freepik<\/p>\n

Freepik makes no claim that F Lite produces images superior to leading image generators like Midjourney\u2019s V7, Black Forest Labs\u2019 Flux family, or others. The goal was to make a model openly available so that developers could tailor and improve it, according to the company.<\/p>\n

That being said, running F Lite is no easy feat. The model requires a GPU with at least 24GB of VRAM.<\/p>\n

Other companies developing media-generating models on licensed data include Adobe<\/a>, Bria<\/a>, Getty Images<\/a>, Moonvalley<\/a>, and Shutterstock. Depending on how AI copyright lawsuits shake out, the market could grow exponentially.<\/p>\n