Sigma Schedule Weighted Mean 🎭🅐🅓

Displayed as Sigma Schedule Weighted Mean, this combines two SIGMA_SCHEDULE objects differently from its sibling ADE_SigmaScheduleSplitAndCombine: instead of switching from one schedule to the other at a chosen point, this blends both curves proportionally across the entire schedule.

How it differs from Split & Combine

Split & Combine gives you a hard switch - schedule A for the first stretch, schedule B for the rest. This node gives you one continuous curve that's a weighted mix of both, the whole way through. weight_A at 1.0 is pure schedule A, 0.0 is pure schedule B, and values in between mix the two proportionally at every point in the schedule, not just at one split index. Use this when you want a noise curve genuinely "between" two known schedules - for instance, splitting the difference between two motion models' recommended beta_schedule curves - rather than one taking over from the other partway through.

This is deep, largely undocumented territory even within an already niche corner of the pack. Neither the README nor community discussion covers what a blended noise curve produces in practice; treat any result here as an experiment, and compare against each pure schedule on its own before trusting the blend.

Inputs and output

All required:

• schedule_A, schedule_B (SIGMA_SCHEDULE) - the two schedules to blend.
• weight_A (default 0.5, range 0–1, step 0.001) - the mix ratio.

Output is a single SIGMA_SCHEDULE, feeding ADE_SigmaScheduleToSigmas for a concrete SIGMAS list, or the sigma_schedule slot on Sample Settings.

How to install it

ComfyUI Manager: search AnimateDiff Evolved, confirm the author is Kosinkadink, install, restart. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved

No model download - pure schedule math on objects built elsewhere in your graph, typically with ADE_RawSigmaSchedule.

Common issues & troubleshooting

Unsure what a blended schedule should even sound or look like. There's no documented reference case to compare against - this is genuinely exploratory. Render short test clips at a few different weight_A values and compare them against the two pure schedules to build your own intuition before committing to a longer run.

Expected a smooth transition mid-generation and got a static blend instead. That's this node's actual behavior - the mix ratio is fixed for the whole schedule. If you want the weighting itself to shift over the course of sampling (favoring A early, B late, say), use ADE_SigmaScheduleWeightedAverageInterp instead, which ramps between a start and end weight.

Result looks broken or degraded, not just "different." Check that both schedule_A and schedule_B are genuinely compatible with the model you're sampling - blending two schedules doesn't average out incompatibility with your checkpoint's actual training if one of them is already the wrong family for it.

Wondering if you need this. Almost certainly not for typical work - picking one named beta_schedule preset on your loader covers nearly every AnimateDiff workflow. This is for deliberate, custom-sampler experimentation.