Skip to main content
The CLIPTextEncodeHiDream node processes four separate text inputs using different language models (CLIP-L, CLIP-G, T5-XXL, and LLaMA) and combines them into a single conditioning output. It tokenizes each text input with its corresponding model and encodes them together using a scheduled encoding approach, enabling more sophisticated text conditioning by leveraging multiple language models simultaneously.

​
Inputs

ParameterDescriptionData TypeRequiredRange
clipThe CLIP model used for tokenization and encodingCLIPYes-
clip_lText input for CLIP-L model processing. Supports multiline text and dynamic prompts.STRINGYes-
clip_gText input for CLIP-G model processing. Supports multiline text and dynamic prompts.STRINGYes-
t5xxlText input for T5-XXL model processing. Supports multiline text and dynamic prompts.STRINGYes-
llamaText input for LLaMA model processing. Supports multiline text and dynamic prompts.STRINGYes-
Note: All four text inputs (clip_l, clip_g, t5xxl, and llama) are required for proper functioning, as each contributes to the final conditioning output through the scheduled encoding process.

​
Outputs

Output NameDescriptionData Type
CONDITIONINGThe combined conditioning output from all processed text inputs, encoded using the scheduled encoding methodCONDITIONING
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): 4d0febcfc7fa8035d817a6b36ec80b9bdf176ede8446b49f291efd96b9d1371c