Fooocus Playground 2.5: New Feature Overview
In this video, Kleebz from Kleebz Tech AI introduces a new feature in Fooocus, specifically the Playground 2.5 model, which was added in the latest update 2.4.3. This model is now integrated into Fooocus, and users do not need to download it separately. Kleebz explains how to enable Playground 2.5 by selecting the appropriate preset in Fooocus, which configures necessary settings automatically, including the scheduler and guidance scale.
Kleebz discusses the optimal settings for using Playground 2.5, suggesting a guidance scale of around 2 and adjusting the number of steps to 40 for better results. He compares images generated with different step settings, showing that 40 steps often produce quality results similar to those generated with 60 steps.
The video highlights the strengths of Playground 2.5, such as enhanced color and contrast, and its limitations, particularly with Loras and image prompts. Kleebz demonstrates how Loras might produce unexpected results because they are not trained on Playground 2.5. He also mentions that image prompts do not influence the generation as expected.
For inpainting, Kleebz provides a workaround by changing the inpaint image setting to ‘none’ in the advanced developer debug mode, which improves the results. He also touches on upscaling, noting that while Playground 2.5 can upscale images, users might sometimes encounter issues and may need to switch to a different model for better results.
Finally, Kleebz compares images generated by Playground 2.5 with those from other models like Juggernaut and Zavy Chroma, showcasing the distinct artistic style and enhanced color contrast of Playground 2.5. He concludes by encouraging viewers to explore the new feature and announces upcoming content on Stable Diffusion 3.