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 Hunyuan3Dv2Conditioning node processes CLIP vision output to generate conditioning data for video models. It extracts the last hidden state embeddings from the vision output and creates both positive and negative conditioning pairs. The positive conditioning uses the actual embeddings while the negative conditioning uses zero-valued embeddings of the same shape.

Inputs

ParameterData TypeRequiredRangeDescription
clip_vision_outputCLIP_VISION_OUTPUTYes-The output from a CLIP vision model containing visual embeddings

Outputs

Output NameData TypeDescription
positiveCONDITIONINGPositive conditioning data containing the CLIP vision embeddings
negativeCONDITIONINGNegative conditioning data containing zero-valued embeddings matching the positive embeddings shape