This is a fine-tune of the SAM2.1 Base+ model, designed for anime. Compared to the original model, it may do better on some anime, worse on others. It can often select an entire character with a single point. It works best with a single object, multiple objects can get parts jumbled up when they cross in front of each other. I hope to release an improved version eventually.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support