CLIPLoader - ComfyUI Built-in Node Documentation - ComfyUI

The CLIPLoader node loads a text encoder model (CLIP, T5, or similar) from a file, making it available for use in other nodes that need to convert text prompts into numerical representations. It supports a wide variety of model architectures, each requiring a specific encoder type.

Inputs

Parameter	Description	Data Type	Required	Range
`clip_name`	The filename of the text encoder model to load. This must be a file located in the `ComfyUI/models/text_encoders/` directory.	STRING	Yes	List of files found in the `text_encoders` folder
`type`	The architecture type of the model being loaded. This determines which specific encoder variant to use (default: `"stable_diffusion"`).	COMBO	Yes	`"stable_diffusion"` `"stable_cascade"` `"sd3"` `"stable_audio"` `"mochi"` `"ltxv"` `"pixart"` `"cosmos"` `"lumina2"` `"wan"` `"hidream"` `"chroma"` `"ace"` `"omnigen2"` `"qwen_image"` `"hunyuan_image"` `"flux2"` `"ovis"` `"longcat_image"` `"cogvideox"` `"lens"` `"pixeldit"` `"ideogram4"` `"boogu"` `"krea2"` `"joyimage"` `"mage"` `"minimax"` `"yue2"` `"flux3"` `"moss_image"`
`device`	The device to load the model onto. `"default"` uses the GPU if available, while `"cpu"` forces CPU loading. This is an advanced option (default: `"default"`).	COMBO	No	`"default"` `"cpu"`

Supported Type-to-Encoder Mappings

The type parameter selects the correct encoder for a given model architecture. The following are common mappings:

Type	Encoder
stable_diffusion	clip-l
stable_cascade	clip-g
sd3	t5 xxl / clip-g / clip-l
stable_audio	t5 base
mochi	t5 xxl
cogvideox	t5 xxl (226-token padding)
cosmos	old t5 xxl
lumina2	gemma 2 2B
wan	umt5 xxl
hidream	llama-3.1 (recommended) or t5
omnigen2	qwen vl 2.5 3B
joyimage	qwen3-vl 8B
lens	gpt-oss-20b
pixeldit	gemma 2 2B elm
minimax	MiniMax H3 Qwen3-VL or Music3 Qwen/RVQ
flux3	qwen3-vl 4B

Outputs

Output Name	Description	Data Type
`CLIP`	The loaded text encoder model, ready to be connected to other nodes for text encoding and conditioning.	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): 3abcb39dd6ce4935f51127d4d38a8474ecf80a4df70d702f3eea68d400be91a5

⌘I