SeedVR2Conditioning - ComfyUI Built-in Node Documentation

This node builds positive and negative conditioning from a VAE latent for use with the SeedVR2 model.

This node builds positive and negative conditioning from a VAE latent for use with the SeedVR2 model. It validates the input latent shape and the model structure, then produces both positive and negative conditioning that guide image or video sampling.

Inputs

Parameter	Description	Data Type	Required	Range
model	The SeedVR2 model.	MODEL	Yes	-
vae_conditioning	The SeedVR2 VAE latent to build conditioning from (display name: latent).	LATENT	Yes	-

Note: The vae_conditioning latent must be a 5-D tensor in Comfy channel-first layout (B, C, T, H, W), where C is the expected SeedVR2 VAE channel count. The node raises an error if the latent is not 5-D, if its channel count does not match, or if it appears to be in channel-last layout. The model input must be a model with the expected SeedVR2 structure. Internally, the node appends a constant mask channel to the latent and attaches the resulting condition to both the positive and negative conditioning pools.

Outputs

Output Name	Description	Data Type
positive	The positive conditioning for sampling.	CONDITIONING
negative	The negative conditioning for sampling.	CONDITIONING

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): 28e508bdd776e2e3f5f2f93bfc29a1a1d1c34a11dbdc7f421d197ddbfa85f0f5