Sigmas Selector (Texturaizer)

ComfyUI's newer sampling path is built on sigma schedules: instead of a sampler secretly deciding how noise is scheduled, you compute a SIGMAS object explicitly and hand it to a custom sampler. Sigmas Selector is that schedule generator for the Texturaizer pack - a small node that takes a model, a scheduler, steps, and denoise, and returns a SIGMAS object ready for a KSamplerSelect / SamplerCustomAdvanced-style flow.

Context: Texturaizer is a paid Blender addon by Luke Kratsios (LatentSpaceDirective) whose free ComfyUI node pack reads its exported JSON. This node doesn't read JSON at all - it's a pure utility, and honestly the most reusable one in the pack outside the Texturaizer workflow itself.

The inputs

• model (MODEL) - the model the schedule is computed for. Schedulers derive their sigma range from the model's noise/denoise range, so this isn't optional decoration.
• scheduler - stock ComfyUI schedulers (simple, karras, exponential, sgm_uniform, ddim_uniform, beta, normal, …) plus the extras this pack adds: AYS SDXL, AYS SD1, AYS SVD, and GITS[coeff=1.2].
• steps (INT) - how many sampling steps the schedule spans.
• denoise (FLOAT, 0–1) - the schedule's tail. 1.0 means full denoise; lower values give you a truncated schedule for img2img-style work.

Output: SIGMAS, which plugs into the sigma input of custom sampler nodes.

How it works

The node is a thin but interesting wrapper. For stock schedulers it delegates to ComfyUI's BasicScheduler. But when the scheduler starts with AYS, it routes to ComfyUI's AlignYourStepsScheduler (AYS SDXL → its SDXL preset, AYS SD1 → SD1, AYS SVD → video); when it starts with GITS[coeff=, it routes to GITSScheduler with the coefficient parsed out. So the "extra" scheduler names in the dropdown aren't fake - they're real sigma-optimization schedules (Align Your Steps is the "fewer steps, same quality" trick) that stock KSampler Basic wouldn't know how to run.

That's genuinely the reason this node exists: Texturaizer's sampler data can name AYS or GITS schedulers, and this selector is what lets those names actually execute. If you only ever use karras or normal, it's a plain BasicScheduler with extra steps.

Install

Pack-level, once:

cd ComfyUI/custom_nodes
git clone https://github.com/LatentSpaceDirective/ComfyUI-Texturaizer

or search "Texturaizer" in ComfyUI Manager, restart. Stock ComfyUI dependencies only, no models to fetch.

When to reach for it

Two real scenarios. In a Texturaizer graph, when Get Sampler Data feeds the pack's sampler, this node sits between them if you want the advanced AYS/GITS schedules - the sampler node's own scheduler dropdown can't reach them the same way. Outside Texturaizer, it's a decent standalone BasicScheduler substitute whenever you're building a custom-sampler chain and want the AYS presets without hunting for the core nodes. Just remember the output is a schedule, not a sample - pair it with something that consumes SIGMAS, or nothing happens.