ComfyUI Node: Egregora RNNoise Denoise

Egregora RNNoise Denoise:

The Egregora_RNNoise_Denoise node is designed to enhance audio quality by reducing noise, specifically focusing on speech. It operates at a high sample rate of 48 kHz, processing audio in 10 ms chunks, which equates to 480 samples per chunk. This node can handle both mono and stereo audio, either processing each channel separately or downmixing to mono for a unified approach. It leverages the pyrnnoise library, version 0.3.x or higher, utilizing the denoise_chunk API to effectively reduce noise. The node incorporates a static strength parameter, an adaptive mix driven by per-frame Voice Activity Detection (VAD), and a post-gain feature with a ceiling to ensure optimal audio output. This combination of features makes it particularly effective for improving the clarity of speech in audio recordings, making it a valuable tool for AI artists working with audio content.

Egregora RNNoise Denoise Input Parameters:

x48_np

This parameter represents the input audio data that the node will process. It should be a NumPy array containing audio samples at a 48 kHz sample rate. The audio data is expected to be in a normalized floating-point format, where the values range between -1.0 and 1.0. The node will convert these values to 16-bit integers for processing. The quality of the input audio can significantly impact the effectiveness of the denoising process, so it is recommended to provide clean and well-recorded audio samples for optimal results.

Egregora RNNoise Denoise Output Parameters:

vad_probs_per_frame

This output parameter provides the Voice Activity Detection (VAD) probabilities for each frame of the processed audio. It is a NumPy array of floating-point values, where each value represents the likelihood of speech presence in the corresponding audio frame. These probabilities can be used to understand the speech dynamics within the audio, offering insights into which parts of the audio contain speech and which are likely to be noise. This information can be particularly useful for further audio processing or analysis tasks.

Egregora RNNoise Denoise Usage Tips:

• Ensure that your input audio is sampled at 48 kHz to match the node's processing requirements. This will help maintain the quality and accuracy of the denoising process.
• If working with stereo audio, consider whether you want to process each channel separately or downmix to mono, depending on your specific needs and the nature of the audio content.

Egregora RNNoise Denoise Common Errors and Solutions:

RNNoise not installed

• Explanation: This error occurs when the pyrnnoise library is not installed in your environment, which is necessary for the node to function.
• Solution: Install the pyrnnoise library by running pip install pyrnnoise in your command line or terminal.

AttributeError: 'RNNoise' object has no attribute 'denoise_chunk'

• Explanation: This error suggests that the installed version of pyrnnoise does not support the denoise_chunk API, which is required by the node.
• Solution: Ensure that you have pyrnnoise version 0.3.x or higher installed. You can upgrade the library by running pip install --upgrade pyrnnoise.

ValueError: Input audio must be a NumPy array

• Explanation: This error indicates that the input provided to the node is not in the expected NumPy array format.
• Solution: Convert your audio data to a NumPy array with a sample rate of 48 kHz before passing it to the node. Use libraries like numpy and scipy to handle audio data conversion if necessary.