Key Takeaways
- Freepik, a graphic design platform, has launched a new AI image generator named F Lite.
- Unlike many AI models, F Lite was trained exclusively on commercially licensed, work-safe images, aiming to avoid copyright problems.
- The model is “open,” allowing developers to freely use, adapt, and improve it.
- It joins a growing number of AI tools being developed using licensed data instead of content scraped from the wider web.
Online graphic design platform Freepik has introduced a new AI tool that creates images from text descriptions. Called F Lite, it stands out because it was built using only licensed images that are safe for work.
This approach tackles a major headache in the AI world: copyright. Many AI image generators learn by processing vast amounts of online content, often including copyrighted material without permission. This has led to lawsuits against companies like OpenAI and Midjourney.
Freepik developed F Lite with AI startup Fal.ai, training the system on about 80 million images from Freepik’s own collection, according to TechCrunch. They used powerful computer hardware over two months to build the model.
By using licensed data, Freepik aims to provide a legally safer option. F Lite comes in two styles: a standard version that closely follows text prompts and a “texture” version designed for more creative, though sometimes less predictable, results.
Freepik isn’t claiming F Lite produces better images than top competitors. Instead, their goal was to release an open model that developers could experiment with and build upon.
Using F Lite does require some serious computing power, needing a graphics card (GPU) with at least 24GB of memory.
Freepik joins other companies like Adobe, Getty Images, and Shutterstock who are also developing AI tools trained on properly licensed content. As legal battles over AI and copyright continue, models trained this way could become increasingly important.
