ComfyUI Node: FL YuE2 · Decode Audio

FL YuE2 · Decode Audio:

The FL_YuE2_Decode node is designed to transform encoded music data into high-quality audio output. It specifically decodes music latents into 48 kHz stereo audio, making it suitable for professional audio applications. This node is integral to the audio processing pipeline, allowing you to convert abstract music representations into tangible audio files that can be previewed or saved. The decoding process respects the original boundaries of the audio tiles, ensuring that the output maintains its intended structure without unwanted crossfades. This feature is particularly beneficial for preserving the integrity of the audio during the decoding process.

FL YuE2 · Decode Audio Input Parameters:

audio_decoder

The audio_decoder parameter requires a YUE2_VAE type input, which is responsible for handling the decoding process. This component is crucial as it interprets the encoded music data and translates it into audio. The effectiveness of the decoding process heavily relies on the capabilities of the audio decoder, ensuring that the output audio is of high quality and fidelity.

music_latents

The music_latents parameter accepts YUE2_LATENTS, which are the encoded representations of the music. These latents contain the essential information needed to reconstruct the audio. The quality and characteristics of the final audio output are directly influenced by the data contained within these latents, making them a critical component of the decoding process.

tile_frames

The tile_frames parameter is an integer that determines the size of the audio tiles during decoding, with a default value of 1024. It can range from a minimum of 64 to a maximum of 2048, adjustable in steps of 64. Smaller tile sizes can reduce the VRAM usage of the decoder, which is beneficial for systems with limited resources. However, using smaller tiles may also impact the processing time and the smoothness of the audio output. The default value of 1024 is recommended for most scenarios as it balances performance and resource usage effectively.

FL YuE2 · Decode Audio Output Parameters:

audio

The audio output parameter provides the decoded audio in a format that can be directly used for playback or further processing. This output is a stereo waveform at a sample rate of 48 kHz, ensuring high-quality sound reproduction. The audio output is crucial for any application that requires listening to or analyzing the decoded music, as it represents the final product of the decoding process.

FL YuE2 · Decode Audio Usage Tips:

• To optimize performance, adjust the tile_frames parameter based on your system's VRAM capacity. Lower values can help reduce memory usage but may increase processing time.
• Ensure that the music_latents provided are correctly formatted and contain the necessary data for accurate audio reconstruction.
• Use the default tile_frames value of 1024 for a balanced approach between performance and resource usage unless specific conditions require adjustments.

FL YuE2 · Decode Audio Common Errors and Solutions:

Expected YuE2 latents shaped [batch,64,frames]

• Explanation: This error occurs when the music_latents input does not have the expected shape, which should be a three-dimensional array with specific dimensions.
• Solution: Verify that the music_latents input is correctly formatted and matches the required shape. Ensure that the data is properly pre-processed before being fed into the node.

YuE2 produced non-finite audio

• Explanation: This error indicates that the decoding process resulted in audio data that contains non-finite values, which can occur due to issues in the decoding process or corrupted input data.
• Solution: Check the integrity of the music_latents and audio_decoder inputs. Ensure that they are not corrupted and are compatible with the node's requirements. If the problem persists, consider re-encoding the music data.