Freepik has introduced F Lite, an AI-powered text-to-image model that puts copyright safety first. This innovative tool is built exclusively using Freepik’s commercially licensed images, a move that sets it apart from other models grappling with copyright issues. With around 10 billion parameters under its hood, F Lite represents a fresh take on generating illustrations and vector-style images in a secure environment.
Developed in partnership with Fal.ai, F Lite was trained using 64 Nvidia H100 GPUs over two months on an internal dataset of about 80 million images. This intensive process ensures that the model is optimised for creating images while respecting copyright laws. It’s a practical alternative at a time when other companies face legal scrutiny over using publicly scraped images.
Available as an open-source project on GitHub, F Lite does require a GPU with at least 24 GB of VRAM, though you can also try out a demo on Hugging Face. While the model shines with more detailed prompts and vector-style imagery, it can struggle with photorealistic outputs or shorter descriptions and still has some difficulty with integrating text within images. If you’ve ever spent time tweaking outputs, you might appreciate Freepik’s honest appraisal of its model’s strengths and limitations.
Based in Spain, Freepik continues to serve creative professionals with a vast library of digital graphics, stock photos, and templates. The company is also expanding its portfolio with AI-driven tools for automatic image editing and synthetic image generation. For designers looking for a reliable, copyright-safe solution, F Lite could be just the tool to simplify your workflow.
