Core gameplay: Everything can be "bodified" (Cosplay Capable). Thanks to Qwen's powerful semantic understanding, you can make this Weird dolls play any role: from Spider - Man on a rainy night to the Dragon Prince of the Tang Dynasty; from the Corpse Bride in the Victorian era to the cyberpunk mechanical monk. It retains the clothing characteristics of the role, but will forcefully replace the material with Weird dolls, creating a unique atmosphere of Chinese Gothic and B - Movie Horror.
Usage cases:
Two weird dolls portraying Donald Trump and Hillary Clinton dancing the waltz in an empty White House hallway. They both smile stiffly at the camera. The image is a low-angle full shot.
A weird doll portraying Albert Einstein riding a small bicycle on an abandoned street. He is sticking his tongue out at the camera. The image is a full body shot .
Two weird dolls portraying Elvis Presley and Marilyn Monroe standing side by side against an abandoned stage. Elvis wears a white jeweled jumpsuit with a high collar, and Marilyn holds down her billowing white dress. They strike a pose and look at the camera. The image is a full body shot.
4. Two weird dolls portraying The Joker and Harley Quinn dancing waltz against inside an abandoned subway. They both smile manically at the camera. The image captures a full body shot
--------------------------------------------------------------------
If you find any problems, please look below:
1.Sometimes some IPs are more difficult, and the characteristics of the weird doll may be lost, such as the lack of a big belly. You can draw a card or input "xx has a big belly." Both can improve the probability. That is, writing out its characteristics can increase the success rate.
2. It is recommended that the batch size be 4 or more because there is a probability of drawing cards.
3.If you think it's too slow, you can try another application of mine, which uses the zimage model. The Qwen version has higher color saturation, but its aesthetic feeling may be weaker than that of zimage. However, its migration ability is stronger and can generate images with skin color closer to the original IP. Each has its own merits. It is recommended that you can use both. If the IP generated by one does not meet your imagination, the other will most likely be very good.
------------------
The address of my other LoRA -- theZimage version