The Lora model can now generate Lego toys with packaging in the background.
It's even capable of crafting the text on the packaging.
However, the composition of the image is quite random, so it might take several attempts to achieve a desirable composition.
The trigger word is 'Lego', and prompts like 'Packaging' or 'Box' can help ensure the packaging is also depicted.
If anyone has ideas on how to consistently improve the composition, please let me know in the comments.