Inputs
Note: The
mask input is required and must be provided. If the mask has a different batch size than the image, it is automatically repeated to match. If the mask dimensions differ from the image dimensions, the mask is resized to match the image using bilinear interpolation. The output size is automatically rounded down to the nearest multiple of 16 to ensure compatibility with DINOv3 patch and Flux2 VAE stride requirements.
Outputs
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub
Source fingerprint (SHA-256):
3f33dbc3a99ccb23ede767915a28fabdfa388edb8d5782edea3f8d03e5965b2a