GR Sigma Presets

A quick primer if SIGMAS is a type you haven't touched: it's the noise schedule a sampler consumes - how much noise gets removed at each step, as distinct from the sampler algorithm itself. ComfyUI splits the two on purpose, and a SIGMAS output is what you feed into a custom-sampler chain (SamplerCustom, SamplerCustomAdvanced) instead of letting a stock KSampler pick a scheduler for you. This pack ships four different nodes for building or picking one; this is the simplest of the four - one dropdown, ten fixed curves, nothing else to configure.

How it works

Pick a preset and you get a SIGMAS output, full stop. Look at the names and the intent is obvious: this isn't a general txt2img scheduler picker, it's tuned for image-to-image and video-continuation work, where "how much should this get changed" is the actual knob you're turning. Ultra Lock (0.0 only) – No change, structure frozen is a zero curve - a genuine no-op, useful as a reference point or a deliberate pass-through rather than something you'd pick to actually generate with. From there the list runs Micro Detail, Low Motion, Img2Img Safe, Balanced I2V, Stylised Motion, a manual preset, and a few structure-biased and wide-tail options for more aggressive changes.

The thing to notice: there's no steps input here at all. Each preset arrives at a fixed length already baked in - this node doesn't rescale to however many steps you're actually sampling at. That's the real difference from its sibling GR Sigma Preset Selector, which takes the same kind of idea but adds a steps input and auto-scales the curve to fit. If none of these ten curves lines up with your step count, that sibling (or GR Sigmas for full manual control) is the better fit.

The inputs and outputs that matter

• preset - the only input, one of ten fixed curves.

One output: sigmas (SIGMAS) - wire it directly into your custom sampler's sigma input.

How to install it

ComfyUI Manager, search GraftingRayman. Manual:

cd ComfyUI/custom_nodes
git clone https://github.com/GraftingRayman/ComfyUI_GraftingRayman

Restart afterward. The whole pack fails to import without OpenAI's CLIP installed separately, even though this node never touches CLIP - portable ComfyUI: .\python_embeded\python.exe -m pip install git+https://github.com/openai/CLIP.git; system Python: pip install git+https://github.com/openai/CLIP.git.

Common issues & troubleshooting

Pack won't load. Check your startup console for the CLIP import error and run the pip install above.

Nothing here matches my step count. By design - there's no steps or auto-scale option on this node. Switch to GR Sigma Preset Selector for the same idea with step-count rescaling, or GR Sigmas if you want to hand-build the curve.

Result barely changes the image, or wipes it entirely. Check you didn't leave Ultra Lock selected by accident - that preset is a literal zero curve, and depending on how your sampler chain handles a flat 0.0 schedule, that's either "no change at all" or effectively no denoising happening.

Not sure this is the right node for a text-to-image generation. It probably isn't - the preset names skew toward img2img and video-continuation strength curves, not general-purpose txt2img scheduling. GR Sigma Preset Selector's subject-matched presets (portrait, landscape, and so on) are the better starting point there.