☁️BizyAir FlipSigmas

This is a one-job node: take a sigma schedule and reverse its order. That sounds almost too small to write about, except reversed sigmas are the building block behind unsampling and several of the more elaborate detail-boosting recipes people build around Detail Daemon - the kind of workflow where you need denoising to run in the opposite direction from normal.

How it works

A scheduler normally hands you sigmas descending from high noise to low - that's a standard forward denoising run. Flipping that order is what lets you push a real image back up through the noise schedule instead of down, which is the core move behind "unsampling" or renoise-based workflows: take a latent that already looks like something, walk it partway back toward noise using the reversed schedule, then resample it forward again for a controlled variation. It also shows up as a component people plug into more advanced sigma chains alongside Detail Daemon nodes - one user described their setup as "the detail daemon sampler and lying sigma in succession... plugged a custom scheduler into the sigma node for the CustomSamplerAdvanced" while chasing sharper texture on Flux and HiDream renders.

The inputs and outputs that matter

• sigmas - the only input. Whatever schedule your scheduler node produced.
• Output: SIGMAS - the same schedule, reversed.

That's the entire interface. No tunable parameters, because there's nothing to tune - it's a pure order-reversal operation.

How to install it

ComfyUI Manager: search BizyAir, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/siliconflow/BizyAir

then restart. BizyAir wants an account and API key set up on first use per the README, though a pure tensor-reversal op like this one has no obvious reason to need a heavy cloud model behind it once you're past the login step.

Common issues & troubleshooting

Plugging it into a normal generation and getting garbage. This is the most common mistake - flipping a standard descending schedule and feeding it straight to an ordinary sampler expecting the normal order breaks denoising outright. This node is for graphs specifically designed around the reversed direction (unsampling, custom sigma-chain recipes), not a drop-in replacement for your regular scheduler output.

Not sure if you actually need this. If your workflow doesn't call for renoising an existing latent or a specific sigma-chain trick, you probably don't - most standard txt2img/img2img graphs never touch this node at all.

Combining it with Detail Daemon and getting unpredictable results. That combination is a genuine community pattern, but it's fiddly by nature - start from a known-working shared workflow rather than guessing at the ordering and settings from scratch.