Skip to main content
This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! Edit on GitHub
The CosmosImageToVideoLatent node creates video latent representations from input images. It generates a blank video latent and optionally encodes start and/or end images into the beginning and/or end frames of the video sequence. When images are provided, it also creates corresponding noise masks to indicate which parts of the latent should be preserved during generation.

Inputs

ParameterData TypeRequiredRangeDescription
vaeVAEYes-The VAE model used for encoding images into latent space
widthINTNo16 to MAX_RESOLUTIONThe width of the output video in pixels (default: 1280)
heightINTNo16 to MAX_RESOLUTIONThe height of the output video in pixels (default: 704)
lengthINTNo1 to MAX_RESOLUTIONThe number of frames in the video sequence (default: 121)
batch_sizeINTNo1 to 4096The number of latent batches to generate (default: 1)
start_imageIMAGENo-Optional image to encode at the beginning of the video sequence
end_imageIMAGENo-Optional image to encode at the end of the video sequence
Note: When neither start_image nor end_image are provided, the node returns a blank latent without any noise mask. When either image is provided, the corresponding sections of the latent are encoded and masked accordingly.

Outputs

Output NameData TypeDescription
latentLATENTThe generated video latent representation with optional encoded images and corresponding noise masks