Sampler Selection
Define your sampler + scheduler once, ship it through a pipe
- pipe
Here's the "Sampler Selector's older sibling with a scheduler strapped to it." Instead of outputting just a sampler name, Sampler Selection bundles your sampler + scheduler choice into a pipe you can pass around a whole workflow. Set it once, and every node that reads the pipe knows exactly how you want denoising done - no more hunting through widgets when you decide DPM++ 2M Karras should be euler instead.
How it works
Two dropdowns, one output:
- Sampler - the full 44-entry ComfyUI sampler list (euler, euler_ancestral, dpmpp_2m, lcm, ddim, and so on).
- Scheduler - the noise-schedule list: simple, sgm_uniform, karras, exponential, ddim_uniform, plus the extras the RvTools author appends -
AYS SDXL,AYS SD1,AYS SVD,GITS[coeff=1.2],OSS FLUX,OSS Wan,simple_test.
Pick both and the node packs them into a pipe - internally it's a list shaped as [sampler, scheduler, 0, 0.0]. Those trailing placeholders are kept deliberately so the pipe stays compatible with the pack's smaller sampler-settings pipe nodes. Feed this pipe into anything that consumes the RvTools sampler-settings pipe format, and your sampler and scheduler ride along as one unit.
Why it matters
The scheduler is the half of the equation people keep forgetting, and it's often the difference between "looks right" and "looks like a smear." Quick grounding: sampler is the algorithm that removes noise step by step; scheduler decides how much noise comes off each step. Karras concentrates denoising in the middle steps, which is why "DPM++ 2M Karras" is the classic SD1.5/SDXL pairing. By bundling both into one pipe, this node makes sure the two never drift apart - you change them together or not at all.
One honest caveat from the wider ecosystem: the "Karras is always better" reflex is a DDPM-era habit. Flow-matching models (Flux and the 2025-26 crop) often prefer simpler schedules and low step counts, so treat the presets here as a starting point rather than gospel - the node gives you the whole list precisely so you can experiment.
The input and output
Just the two dropdowns in, and the single pipe out. There's no widget-forcing, no state - it's a pure "define the setting, ship it" node. Wire the pipe output into the RvTools pipe consumers (sampler-settings pipes, Pipe In Context's sampler/scheduler inputs if you adapt them, or the Save Images generation-data path).
Installing it
Part of the RvTools v2 pack - no models to download. ComfyUI Manager → search RvTools → install ComfyUI-RvTools_v2, restart. Or:
cd ComfyUI/custom_nodes
git clone https://github.com/r-vage/ComfyUI-RvTools_v2
Restart and check the console for RvTools v2: Version 2.5.x. Extra deps (opencv-python, pilgram, pynvml, piexif) come through Manager automatically.
Troubleshooting
- Custom schedulers are part of the list, but they come from other packs. The AYS/OSS entries are appended by RvTools itself; if a workflow expects a scheduler you don't see, check what schedule packs are installed.
- The pipe won't plug into arbitrary packs. This is RvTools' own pipe format. It connects to other RvTools pipe nodes cleanly; for other packs you'll need an adapter.
- Old v1 RvTools references fail to install - the original repo was deleted, which is why v2 renamed everything. The author's current project, ComfyUI_Eclipse, supersedes this pack.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| Sampler | COMBO | 44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38 | |
| Scheduler | COMBO | 16 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +10 |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| pipe | pipe | — |