LTXVEmptyLatentAudio - ComfyUI Built-in Node Documentation

The LTXV Empty Latent Audio node creates a batch of empty (zero-filled) latent audio tensors.

The LTXV Empty Latent Audio node creates a batch of empty (zero-filled) latent audio tensors. It reads the configuration of a connected Audio VAE model to determine the correct latent dimensions, such as the number of channels and frequency bins, and calculates how many audio latents are needed from the frame count and frame rate. The resulting empty latent can be used as a starting point for audio generation or manipulation workflows.

Inputs

Parameter	Description	Data Type	Required	Range
frames_number	Number of frames. The default value is 97.	INT	Yes	1 to 1000
frame_rate	Number of frames per second. The default value is 25.0. This input accepts either FLOAT or INT values.	FLOAT	Yes	1.0 to 1000.0
batch_size	The number of latent audio samples in the batch. The default value is 1.	INT	Yes	1 to 4096
audio_vae	The Audio VAE model to get configuration from. Displayed as “Audio VAE”.	VAE	Yes	N/A

Note: The audio_vae input is mandatory. The node raises an error if it is not provided.

Outputs

Output Name	Description	Data Type
Latent	An empty latent audio tensor with the shape (batch_size, z_channels, num_audio_latents, audio_freq), where the channel count and frequency bins come from the Audio VAE and the number of audio latents is derived from frames_number and frame_rate. The output also includes a type field set to “audio”.	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): 3ac1bf17ebdba7c3a73bdd795f561b7bee31798d8a1efc11b972db1944f873a4