Animate Any 2

Recent Animation Methods of character image based on spraying models, such as life with any, causing significant progress to minimize character animations. However, these methods fail to produce reasonable associations between characters and their surroundings. To answer this limit, we introduce a person of 2, trying to advertise the environment-related characters. More than picking motion signals from origin video, we arrest environmental representations as input conditions. The environment is formed as the region of separating characters and our model produces characters to proceed to these regions as the interaction context continues. We recommend a strategic strategy more effectively indicating the relationship between character and environment. In addition, to enhance honesty with interactions with things, we turned away something to an object to take part in the objects with the spirial mixing for a partial injection . We also identify a behavioral strategy that makes the model management of many different patterns of action. Experimental results reflect the best performance in the proposed approach.
https://humanaigc.github.io/animate-anyone-2/static2/images/motivation.png
2025-02-16 14:20:00