MiniMaxH3FunControlNetApply - ComfyUI Built-in Node Documentation

This node applies a MiniMax H3 Fun ControlNet to a text-to-video model as a model patch. It can use an optional control video and an optional mask to steer generation, and it returns a patched copy of the model for further sampling. When the strength is set to 0, or when neither a control video nor a mask is provided, the input model is returned unchanged.

Inputs

Parameter	Description	Data Type	Required	Range
`model`	The diffusion model to which the MiniMax H3 Fun ControlNet patch is applied.	MODEL	Yes	N/A
`model_patch`	The MiniMax H3 Fun ControlNet patch whose control signals are injected into the model; it must be compatible with the given `model`.	MODEL_PATCH	Yes	N/A
`vae`	VAE used to encode the control and source video frames into the latent space expected by the model.	VAE	Yes	N/A
`strength`	Overall strength of the ControlNet effect. When set to 0, the node does nothing and returns the input model unchanged. (default: 1.0)	FLOAT	Yes	min 0.0, max 10.0, step 0.01
`start_percent`	Start of the sampling range, expressed as a percentage of the sampling schedule, during which the ControlNet is active. It is converted internally to the equivalent sigma value. (default: 0.0)	FLOAT	Yes	min 0.0, max 1.0, step 0.001
`end_percent`	End of the sampling range, expressed as a percentage of the sampling schedule, during which the ControlNet is active. It is converted internally to the equivalent sigma value. (default: 1.0)	FLOAT	Yes	min 0.0, max 1.0, step 0.001
`control_video`	Optional video frames used as the ControlNet visual hint. The frames are resized to match the generated video and encoded with the `vae`.	IMAGE	No	N/A
`mask`	1 marks the regions to regenerate. Mask values above 0.5 are treated as marked regions.	MASK	No	N/A
`source_video`	Video behind the mask; only read when a mask is given.	IMAGE	No	N/A

Note: For the patch to have an effect, strength must be greater than 0 and at least one of control_video or mask must be supplied. source_video is ignored unless mask is provided; if mask is given without source_video, the content behind the masked regions is treated as black.

Outputs

Output Name	Description	Data Type
`model`	A patched clone of the input model with the MiniMax H3 Fun ControlNet applied. If `strength` is 0, or no control video or mask is supplied, the original model is returned unchanged.	MODEL

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): e907fb8e5ae60663d1d10b315985695ee5d49397fef6bd76b0e723637457a74a

​Inputs

​Outputs

Inputs

Outputs