Flex Audio Visualizer Line ⚡🅡🅞🅣🅘

If you need the audio visualizer you've seen a thousand times - a row of bars pulsing to the beat, or a smooth waveform line - this is that node, rendered directly to an image sequence you can composite over anything. It's the "generic" visualizer in the pack; its sibling, FlexAudioVisualizerContour, does the fancier version that traces a mask's shape instead of sitting in a straight line.

How it works

Feed it audio and it analyzes it via FFT, either showing visualization_feature: frequency (a proper spectrum analysis, low to high) or waveform (raw amplitude - more organic, less "EQ display"). visualization_method picks the render style: bar for discrete bars, line for a continuous connected line through the same data. frame_rate controls how many image frames come out, independent of the audio's own sample rate.

Everything else is layout and shape. screen_width/screen_height set the canvas, position_x/position_y place the visualizer's center within it (0–1, so 0.5/0.5 is dead center), and rotation spins the whole thing - useful for a vertical bar stack instead of horizontal. length sets how far the visualization extends before auto-fitting to the screen edges (0 = auto-fit based on rotation). For bars specifically: num_bars, max_height/min_height, separation (gap between bars), and curvature (rounds the bar corners). For the line mode: curve_smoothing. reflect flips the whole render to draw downward from the baseline instead of up - handy for a mirrored, "meter-style" look.

Like every Flex node, it accepts an optional opt_feature that can additionally modulate one of 14+ of the above parameters (feature_param picks which; the schema shows the first dozen - smoothing, rotation, position_y, num_bars, max/min height, separation, curvature, reflect, curve smoothing, fft_size, min/max frequency - plus a couple more), with feature_mode (relative/absolute), feature_threshold, and strength controlling how that modulation behaves. In practice most people leave opt_feature unconnected and just let the audio itself drive the bars via visualization_feature - the feature input is for a second, independent reactive layer on top (e.g. motion from another clip nudging the rotation while the audio drives the bar heights).

Inputs and outputs that matter

The ones worth touching first: audio, visualization_method (bar vs line), num_bars or length, and max_height/min_height to fit your canvas. Outputs are IMAGE (the rendered visualization) and MASK (the same shape as a mask, so you can composite it, recolor it, or use it to punch a hole in something else).

How to install it

ComfyUI Manager, search "RyanOnTheInside," or:

cd ComfyUI/custom_nodes
git clone https://github.com/ryanontheinside/ComfyUI_RyanOnTheInside
cd ComfyUI_RyanOnTheInside
pip install -r requirements.txt

then restart. Audio analysis depends on librosa; make sure that install step actually finishes without errors before you go hunting for bugs elsewhere.

Common issues & troubleshooting

Bars/line render but never move. Check visualization_feature is set to something meaningful for your source - silent or near-silent audio produces a flat visualization regardless of settings, since there's nothing in the signal to display.

Output frame count doesn't match your other video nodes. frame_rate here is independent from your project's video frame rate; if you're compositing this over footage, set frame_rate to match, or run the result through a frame-rate/interpolation node afterward.

It looks cramped or clipped. screen_width/screen_height define the canvas this node renders at - if you composite the output onto a differently-sized image without resizing, bars can overflow or look tiny. Match the canvas to your target resolution up front rather than scaling after the fact.