ComfyUI Extension: AudioDriven-Latent-Space-Tools-for-ComfyUI
This is a custom node for ComfyUI that analyzes audio files using Librosa, extracting tempo, beat times, energy levels, and timestamps. The analysis results can be displayed in a text box within ComfyUI.
Custom Nodes (0)
README
AudioDriven Latent Space Tools for ComfyUI
Generate dynamic latent noise patterns driven by audio analysis.
Nodes
Librosa Audio Analysis
Analyzes audio files using various methods:
- Onset detection (note starts)
- Spectral content
- Beat tracking
- Tempo analysis
- MEL spectrograms
Audio To Noise Parameters
Converts audio analysis into noise parameters:
- Intensity based on audio energy
- Grain size from spectral features
- Persistence from temporal changes
Audio Noise to Latent
Converts noise parameters to latent space noise using:
- Gaussian noise
- Salt & pepper noise
- Perlin noise
Advanced Audio Noise Patterns
Creates musical visualization-style patterns:
- Simplex noise
- Cellular patterns
- Fractal Brownian Motion (FBM)
- Wave patterns
- Domain warping
Usage
- Input audio file through Librosa Analysis
- Choose analysis type (onset/tempo/mel/etc.)
- Convert to noise parameters
- Generate latent noise
- Connect to ksampler
- Use with low denoising (<0.5) for audio-visual blending
WARNING: Advanced Audio Noise Patterns Node
Typical image counts: Average measurements by analysis type (from 3s, 11.65s, and 89s samples)
Default: huge (~122K for 89s)
Mel/spectral/tempo: high (~7.6K for 89s)
Onset: moderate (60-500 depending on complexity)
Beat: low (14-168 based on tempo)
Second: consistent (~length in seconds)
Half_second: consistent (~2x seconds)
Installation
Download or clone this repository to your ComfyUI\custom_nodes folder
This node is part of a ongoing solo project to integrate music analysis. plz enjoy.