EmptyMiniMaxMusic3LatentAudio - ComfyUI Built-in Node Documentation - ComfyUI

This node creates an empty (zero-filled) latent audio tensor for the MiniMax Music3 model. It converts the requested duration in seconds into the corresponding audio frames and produces a blank latent of the correct size, ready to be used as the starting point for music generation.

Inputs

Parameter	Description	Data Type	Required	Range
`seconds`	The duration of the audio latent in seconds (default: 120.0). The value is converted to audio frames and clamped to the model’s supported duration limits.	FLOAT	Yes	0.04 to model maximum (MAX_AUDIO_FRAMES / AUDIO_FRAMES_PER_SECOND), step 0.04
`batch_size`	The number of audio latents to generate in one batch (default: 1).	INT	Yes	1 to 4096

Note: The seconds value is rounded to the nearest audio frame and clamped to a minimum of 1 frame and a maximum of MAX_AUDIO_FRAMES frames, so the actual latent length may differ slightly from the exact value entered.

Outputs

Output Name	Description	Data Type
`LATENT`	A zero-filled audio latent tensor of shape (batch_size, 128, latent_length). Includes metadata that marks the sample as audio data with a temporal downscale ratio of 512.	LATENT

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): 77e6a69702a837c958c2954bba061c979152f034bc7774a5b6c97ea8d57bda4b