YuE2GenerateMusic - ComfyUI Built-in Node Documentation - ComfyUI

Generates music tokens and acoustic conditioning from a style, lyrics, and an ABC notation. It returns the conditioning and the generated duration in seconds, which should be provided to the Empty YuE2 Latent Audio node. If the ABC input is left empty, the selected mode is ignored and off mode is used automatically.

Inputs

Parameter	Description	Data Type	Required	Range
`clip`	The CLIP model used to tokenize and encode the music inputs.	CLIP	Yes	-
`style`	Text describing the musical style. Supports multiline input and dynamic prompts.	STRING	Yes	Multiline text
`lyrics`	Lyrics for the generated music. Supports multiline input and dynamic prompts.	STRING	Yes	Multiline text
`abc`	Connect the ABC generator or supply an edited score. Leave empty to use off mode automatically. default: ""	STRING	Yes	Multiline text
`seed`	Random seed for generation. default: 0	INT	Yes	0 to 18446744073709551615
`mode`	full: generates melody and chords; melody: generates melody only, recommended for covers. default: “full”	COMBO	Yes	”full" "melody”
`max_duration`	Maximum duration in seconds. Automatically reduced for long prompts; generation can stop earlier. default: 360.0	FLOAT	Yes	0.04 to 900.0
`temperature`	Sampling temperature for generation. default: 1.0 (advanced)	FLOAT	Yes	0.0 to 5.0
`top_p`	Nucleus sampling probability threshold. default: 0.95 (advanced)	FLOAT	Yes	0.01 to 1.0
`top_k`	Top-k sampling limit. default: 100 (advanced)	INT	Yes	1 to 32768
`repetition_penalty`	Penalty applied to repeated tokens. default: 1.2 (advanced)	FLOAT	Yes	0.01 to 10.0

Note: If abc is empty or contains only whitespace, the mode selection is ignored and off mode is used automatically.

Outputs

Output Name	Description	Data Type
`CONDITIONING`	Acoustic conditioning generated from the music tokens.	CONDITIONING
`seconds`	The generated audio duration in seconds. Provide this value to the Empty YuE2 Latent Audio node.	FLOAT

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): 54f5d46cf083726bdf97c86e5683b2727840f75bb553bddf9525cfbf5affa44c