LTXAVTextEncoderLoader - ComfyUI Built-in Node Documentation - ComfyUI

This node loads a specialized text encoder for the LTXV audio model. It combines a text encoder file with a checkpoint file to create a CLIP model used for text conditioning in audio generation. Per the node’s description, the text encoder should be a Gemma 3 12B or a matching Gemma 4 model.

Inputs

Parameter	Description	Data Type	Required	Range
`text_encoder`	The filename of the LTXV text encoder model to load. The available options are loaded from the `text_encoders` folder.	COMBO	Yes	Multiple options available
`ckpt_name`	The filename of the checkpoint to load. The available options are loaded from the `checkpoints` folder.	COMBO	Yes	Multiple options available
`device`	Specifies the device to load the model onto. Use `"cpu"` to force loading onto the CPU. The default behavior (`"default"`) uses the system’s automatic device placement (default: `"default"`). This is an advanced parameter.	COMBO	No	`"default"` `"cpu"`

Note: The text_encoder and ckpt_name parameters work together. The node loads both specified files to create a single, functional CLIP model. The files must be compatible with the LTXV architecture, and the text encoder should be a Gemma 3 12B or matching Gemma 4 model.

Outputs

Output Name	Description	Data Type
`clip`	The loaded LTXV CLIP model, ready to be used for encoding text prompts for audio generation.	CLIP

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): 1f3df2c1791203ba849a87897de14052e0cb8370100dbca19df4cf30169a0a2a