https://civitai.com/user/PatrickStarrrr The training used oil paintings with relatively strong textures. Therefore, it's not recommended to add too many elements or descriptions to the image during the generation process. Otherwise, the combination of objects and textures can make the image appear very cluttered. The example image of the forest actually illustrates this point. It's very suitable for close-ups of people or still lifes, as the background blur + the concrete representation of the subject matter helps to convey the desired atmosphere. The highlights in the portraits are relatively weak. Close-ups of people, still lifes, etc., can also be generated, but the training dataset for portraits was not specifically designed or considered. If using a single Lora, a weight of 0.7-1.0 is recommended. Note: If the generated scene contains a forest, it's recommended to lower the weight. Observations from the uploaded image show that this Lora overfits in forest scenes.