Sigmas PersistentHomology

Persistent homology is a real technique from topological data analysis: you watch a shape evolve and track when its topological features - loops, voids, connected pieces - are "born" and when they "die" as some parameter sweeps through a range. It's genuinely used for things like analyzing the shape of high-dimensional data. RES4LYF doesn't run actual homology computations here; it borrows the vocabulary - birth, death, persistence - to shape a decay curve, using density parameters that stand in for how densely features are born and how densely they die across the schedule.

This sits alongside Sigmas NormalizingFlows, RiemannianFlow, ReactionDiffusion, and ZetaEta as one of RES4LYF's family of mathematically-flavored schedule generators. None of them have an established "use this when" recipe in the community - they exist because they're interesting curve shapes to have available, not because a specific model or workflow calls for them by name. If you came here chasing a known-good schedule, use beta57 or a standard scheduler instead; come back to this one when you're specifically curious what an unusual curve buys you.

How it works

It's a generator - no schedule goes in, only parameters. steps sets the output length; start_value and end_value set the high and low bounds, the same role as max/min sigma anywhere else. persistence_type (linear, exponential, logarithmic, sigmoidal) picks the overall shape of how the curve moves between those bounds. birth_density and death_density (0–1 each, defaulting to 0.3 and 0.7) tune where along the schedule the curve's character shifts - think of them as marking roughly where the "early" and "late" phases of the curve hand off to each other.

The inputs and outputs that matter

All six parameters are required: steps, start_value, end_value, persistence_type, birth_density, death_density. The single output is SIGMAS.

Installing it

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

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

Use the portable build's embedded pip.exe if that's your setup. Restart and hard-refresh afterward.

Where people get tripped up

Because birth_density and death_density are both fractions of the schedule's length, setting birth_density higher than death_density inverts the usual "born early, dies late" ordering the defaults imply - if the resulting curve looks backwards, that's the first thing to check.

Beyond that, treat this the same way as the pack's other generator nodes: verify the output is actually a sane, mostly-decreasing curve (a preview or plot node helps) before wiring it straight into a sampler, since there's no community track record yet to tell you what "normal" looks like for this one.