GR Sigma Preset Selector Advanced

Same idea as its sibling GR Sigma Preset Selector - pick a subject-matched noise curve, tell it your step count, get back sigmas plus a graph - with one addition that earns the "Advanced" in the name: control over the interpolation shape used to spread the preset across your steps, instead of always accepting whatever pacing the preset was designed with.

How it works

preset (26 subject-matched choices, zimage default), steps, and auto_scale_to_steps work identically to the plain Preset Selector - see that article if you need the full rundown on what the presets cover. The new input is curve_shape: preset (the default, meaning "use the preset's own native shape, same as the non-Advanced node"), linear (even change step to step), easein, easeout (slow-then-fast or fast-then-slow ramps), and cosine / s-curve (smooth acceleration-deceleration, the same easing shapes you'd recognize from animation curves, here applied to how aggressively noise gets removed rather than how something moves on screen).

The practical case for reaching past the plain Selector to this one: a preset's overall subject match is right - say portrait gives you the right general balance of structure versus detail - but the pacing feels off, like too much noise removal is crammed into the last couple of steps. Instead of hand-building a curve from scratch in GR Sigmas, you override just the shape and keep everything else about the preset.

One node-level difference worth knowing: this node is flagged as an output node, unlike its plain sibling, which generally means it renders its graph directly on the node body in the app canvas rather than only handing the image downstream - you'll notice a live preview sitting right on the node itself.

The inputs and outputs that matter

• preset, steps, auto_scale_to_steps - same as GR Sigma Preset Selector.
• curve_shape - preset / linear / easein / easeout / cosine / s-curve; leaving it on preset reproduces the plain Selector's behavior exactly.

Three outputs: sigmas (SIGMAS, to your sampler), graph_preview (IMAGE), and preset_info (STRING).

How to install it

ComfyUI Manager, search GraftingRayman. Manual:

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

Restart afterward. The pack won't import at all without OpenAI's CLIP installed separately, even though this node has nothing to do with 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.

Changing curve_shape doesn't seem to do anything. Check it's not still set to preset - that value deliberately reproduces the preset's own native shape, so it needs to be one of the other five options before you'll see a different curve.

Not sure whether to reach for this over the plain Preset Selector. If a preset's subject match is right but the pacing isn't, this is the node. If the subject match itself is wrong, changing curve_shape won't fix that - try a different preset first.

Want to build a curve from scratch rather than reshaping a preset. That's GR Sigmas - full manual control over composition/mid/detail zones, at the cost of a much bigger input list.