Sigmas QuantileNorm

Quantile normalization is a real, well-established statistical technique - it's how you make two datasets share the same distribution of values without caring about their raw shapes, by mapping each one onto the same set of quantiles. This node applies that idea to sigma schedules: instead of just rescaling min/max like most of RES4LYF's other sigma tools, it reshapes the entire distribution of your schedule's values to match a target.

That target can be a clean textbook distribution (uniform, normal, exponential, logistic), or - set target_distribution to custom - another sigma schedule you feed in directly. That second mode is the genuinely useful one for most people: if you like the statistical "feel" of a particular hand-tuned schedule but want your own step count or your own start/end values, QuantileNorm lets you borrow its shape without borrowing its length.

The inputs and outputs that matter

sigmas is the schedule to reshape. target_distribution picks what you're matching against - uniform, normal, exponential, logistic, or custom. num_quantiles (default 100) controls how finely the mapping is computed; more quantiles means a more precise but more expensive match. The optional reference_sigmas input only matters when target_distribution is custom - that's the schedule whose distribution you're matching to. The output is a single SIGMAS.

Compare this to Sigmas Percentile, which is easy to confuse it with: Percentile rescales a range (measure roughly where the bulk of the values sit, stretch that onto a target range) and leaves the shape of the distribution alone. QuantileNorm goes further - it reshapes the entire distribution itself to match the target's, which is a stronger and more disruptive operation. Reach for Percentile when you want to tame outliers without changing the schedule's character; reach for QuantileNorm when you specifically want your schedule's statistical fingerprint to look like someone else's.

Installing it

ComfyUI Manager: search RES4LYF, install, restart. Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/ClownsharkBatwing/RES4LYF/
cd RES4LYF
pip install -r requirements.txt

Swap in the embedded pip.exe on a portable ComfyUI install. Restart and hard-refresh the page after.

Where people get tripped up

Setting target_distribution to custom without wiring anything into reference_sigmas is the obvious first mistake - that input only does something in custom mode, so if you're not seeing the effect you expect, check that the wire is actually connected.

The other thing worth knowing: matching a distribution isn't the same as matching a shape over time. Two schedules can share the exact same distribution of values while looking completely different laid out step-by-step, because quantile matching only cares about how often each value occurs, not where it sits in the sequence. If your goal is literally "make my schedule look like this reference schedule," check the result visually rather than assuming a statistical match guarantees a visual one.