Normalized Amplitude To Float or Int

NormalizedAmplitudeToNumber is AmplitudeToNumber's smarter sibling. Same job - take the pack's normalized amplitude series and hand it to the rest of ComfyUI as plain FLOAT and INT values - but with three knobs that let you reshape the curve while you convert it. That turns what would be a boring pass-through into the node you use when the 0–1 curve isn't quite in the range your target parameter wants.

The trick is a conditional add: values above a threshold get a boost, everything is then clamped to a ceiling. It's a one-node way to lift the quiet half of a curve into usefulness, or push the whole signal into a higher numeric range.

The knobs

• threshold_for_add (0–1, default 0) - values above this get add_to added to them. Values at or below it pass through untouched.
• add_to (default 0) - the amount added above the threshold. Even a small 0.1–0.2 boost on the quiet sections can stop a curve from going dead in the lows.
• add_ceiling (0.1–4, default 1) - the clamp everything is capped at. Raise it (say, to 2) to scale the whole curve into a range that's friendlier to strength-style parameters, while add_to only inflates the part you care about.

Outputs: FLOAT (the reshaped values) and INT (truncated). Both are per-frame arrays, matching how the amplitude series is structured - wire them into array-aware nodes or a scheduler that walks frames, and the pack's example workflow feeds the FLOAT output into FloatArrayToGraph just to preview it.

The gotchas

The INT output truncates, and truncated 0–1 values are all zeros. If your normalized curve is genuinely 0–1, the FLOAT output is the one to use - or pump the curve up with add_to/add_ceiling first and then the ints become meaningful. Two internal details worth knowing rather than debugging later: the node silently maps NaN to 0 and inf to 1 before doing anything, so broken values won't crash it - they'll just quietly become flat. And "add" is strictly additive, not multiplicative: if you want to scale the whole curve, add_ceiling and add_to together get you close, but this node is built for nudging, not remapping.

Install with the pack: ComfyUI Manager search ComfyUI-AudioScheduler, or git clone https://github.com/a1lazydog/ComfyUI-AudioScheduler into custom_nodes and restart. CPU-only, no models.