The LTXV Audio VAE Decode node converts a latent representation of audio back into an audio waveform. It uses a specialized Audio VAE model to perform this decoding process, producing an audio output with a specific sample rate.

## Inputs

| Parameter | Description | Data Type | Required | Range |
| --- | --- | --- | --- | --- |
| `samples` | The latent to be decoded. | LATENT | Yes | N/A |
| `audio_vae` | The Audio VAE model used for decoding the latent. | VAE | Yes | N/A |

**Note:** If the provided latent is nested (contains multiple latents), the node will automatically use the last latent in the sequence for decoding.

## Outputs

| Output Name | Description | Data Type |
| --- | --- | --- |
| `Audio` | The decoded audio waveform and its associated sample rate. The waveform is a tensor moved to the same device as the input latent, and the sample rate is determined by the Audio VAE model. | AUDIO |

> This documentation was AI-generated. If you find any errors or have suggestions for improvement, please feel free to contribute! [Edit on GitHub](https://github.com/Comfy-Org/embedded-docs/blob/main/comfyui_embedded_docs/docs/LTXVAudioVAEDecode/en.md)

---
**Source fingerprint (SHA-256):** `b410b535f8cfc07b2d5a7814d719970ca813172bbbb04bec22a0ce3be04dfa94`